Full text of "The year without a summer? : world climate in 1816"

See other formats

LIBRARY

OF THE

GRAY HERBARIUM

Digitized by

the Internet Archive

in 2014

https://archive.org/details/yearwithoutsumme1992hari

WORLD CLIMATE IN 1816

EDITED BY
C.R. HARINGTON

Canadian
Museum
of Nature

Musee
canadien
de la nature

THE YEAR WITHOUT A SUMMER?

WORLD CLIMATE IN 1816

EDITED BY
C. R. HARINGTON

Above and cover:

Medallion struck in
southern Germany
in memory of the
great famine of 1816-1817.

The inscription reads:
"Great is the distress,

Oh Lord, have pity." CANADIAN MUSEUM

Both faces shown; from Volcano Weather, OF NATURE

The Story of 1816, The Year Without a Summer

by Henry and Elizabeth Stommel OTTAWA, 1992

WAY 0 1 1992

GRAY HERBARIUM

Published by the:

Canadian Museum of Nature
Ottawa, Canada KIP 6P4

Catalogue No. NM95-20/1 1991-E

Available by mail order from:

Canadian Museum of Nature
Direct Mail Section
P.O. Box 3443, Station "D"
Ottawa, Canada KIP 6P4

Public par le :

Musee canadien de la nature
Ottawa, Canada KIP 6P4

N° de catalogue NM95-20/1 1991-E

L'editeur rcmplet les commandes postales
adressees au :

Musee canadien de la nature
Section des commandes postales
CP. 3443, succursale D
Ottawa, Canada KIP 6P4

Printed in Canada Imprime au Canada

ISBN: 0-660-13063-7 ISBN : 0-660-13063-7

Text pages printed on paper
containing recycled fibre.

Les pages du texte sont imprimes
sur un papier contenant
des fibres recycles.

Print of original handwritten copy of Lord Byron's poem "Darkness". Written
at Geneva during 1816 (courtesy of Princeton University Library).

DARKNESS.

I had a dream, which was not all a dream.

The bright sun was extinguish'd, and the stars

Did wander darkling in the eternal space,

Rayless, and pathless, and the icy earth

Swung blind and blackening in the moonless air ;

Morn came and went — and came, and brought no day,

And men forgot their passions in the dread

Of this their desolation ; and all hearts

Were chill'd into a selfish prayer for light :

And they did live by watchfires — and the thrones,

The palaces of crowned kings — the huts,

The habitations of all things which dwell,

Were burnt for beacons ; cities were consumed,

And men were gather'd round their blazing homes

To look once more into each other's face ;

Printed version of Lord Byron's poem "Darkness" (courtesy of Princeton University Library).

CONTENTS

Acknowledgements 5

Introduction 6

C.R. Harington

General 9

Before Tambora: the Sun and Climate, 1790-1830 1 1

John A. Eddy

Eyewitness Account of the Distant Effects of the Tambora Eruption
of April 1815 12
Michael R. Rampino

The Eruption of Tambora in 1815: Environmental Effects and

Eruption Dynamics 16
Haraldur Sigurdsson and Steven Carey

The Possible Effects of the Tambora Eruption in 1815 on Atmospheric
Thermal and Chemical Structure and Surface Climate 46
R.K.R. Vupputuri

Climatic Effects of the 1783 Laki Eruption 58
Charles A. Wood

The Effects of Major Volcanic Eruptions on Canadian Surface Temperatures 78
Walter R. Skinner

Northern Hemisphere 93
North America

Climate in 1816 and 181 1-20 as Reconstructed from Western North American
Tree-Ring Chronologies 97
J.M. Lough

Volcanic Effects on Colorado Plateau Douglas-Fir Tree Rings 115
Malcolm K. Cleaveland

1816 in Perspective: the View from the Northeastern United States 124
William A. Baron

Expansion of Toronto Temperature Time-Series from 1840 to 1778 Using
Various United States and Other Data 145
R.B. Crowe

Climate in Canada, 1809-20: Three Approaches to the Hudson's Bay
Company Archives as an Historical Database 162
Cynthia Wilson

Climatic Change, Droughts and Their Social Impact: Central Canada,
1811 -20, a Classic Example 1 85

Timothy F. Ball

The Year without a Summer: Its Impact on the Fur Trade and History
of Western Canada 196
Timothy F. Ball

The Ecology of a Famine: Northwestern Ontario in 1815-17 203
Roger Suffling and Ron Fritz

The Development and Testing of a Methodology for Extracting Sea-Ice
Data from Ships' Log-Books 218
Marcia Faurer

River Ice and Sea Ice in the Hudson Bay Region during the Second
Decade of the Nineteenth Century 233

A. J.W. Catchpole

The Climate of the Labrador Sea in the Spring and Summer of 1816,
and Comparisons with Modern Analogues 245
John P. Newell

Spatial Patterns of Tree-Growth Anomalies from the North American
Boreal Treeline in the Early 1800s, Including the Year 1816 255
Gordon C. Jacohy, Jr. and Rosanne D'Arrigo

Early Nineteenth-Century Tree-Ring Series from Treeline Sites

in the Middle Canadian Rockies 266

B. H. Luckman and M.E. Colenutt

How Did Treeline White Spruce at Churchill, Manitoba. Respond
to Conditions around 1816? 281
David C. Fayle, Catherine V. Bentley and Peter A. Scott

The Climate of Central Canada and Southwestern Europe Reconstructed
by Combining Various Types of Proxy Data: a Detailed Analysis of
the 1810-20 Period* 291
J. Guiot

Climatic Conditions for the Period Surrounding the Tamhora Signal
in Ice Cores from the Canadian High Arctic Islands

Bea Taylor Alt, David A. Fisher and Roy M. Koerner

Europe (including Iceland)

1816 - a Year without a Summer in Iceland?
A.E.J. Ogilvie

First Essay at Reconstructing the General Atmospheric Circulation
in 1816 and the Early Nineteenth Century
H.H. Lamb

Weather Patterns over Europe in 1816
John Kington

The Climate of Europe during the 1810s with Special Reference to 1816
K.R. Briffa and P.D. Jones

The 1810s in the Baltic Region, 1816 in Particular: Air Temperatures,
Grain Supply and Mortality
J. Neumann

The Years without a Summer in Switzerland: 1628 and 1816
Christian Pfister

Climatic Conditions of 1815 and 1816 from Tree-Ring Analysis in the
Tatra Mountains

Zdzislaw Bednarz and Janina Trepinska

Major Volcanic Eruptions in the Nineteenth and Twentieth Centuries
and Temperatures in Central Europe
Vladimir Bruzek

Asia

Climate over India during the First Quarter of the Nineteenth Century
G.B. Pant, B. Parthasarathy and N.A. Sontakke

Evidence for Anomalous Cold Weather in China 1815-17
Pei-Yuan Zhang, Wei-Chyung Wang and Sultan Hameed

Was There a Colder Summer in China in 1816?
Huang Jiayou

The Reconstructed Position of the Polar Frontal Zone around Japan
in the Summer of 1816
Yasufumi Tsukamura

The Climate of Japan in 1816 as Compared with an Extremely Cool
Summer Climate in 1783

T. Mikami and Y. Tsukamura

Southern Hemisphere

477

Evidence for Changes in Climate and Environment in 1816 as
Recorded in Ice Cores from the Quelccaya Ice Cap, Peru, the

Dunde Ice Cap, China and Siple Station, Antarctica* 479
Lonnie G. Thompson and Ellen Mosley-Thompson

Changes in Southern South American Tree-Ring Chronologies

following Major Volcanic Eruptions between 1750 and 1970 493
Ricardo Villalba and Jose A. Boninsegna

Tree-Ring Chronologies from Endemic Australian and New Zealand
Conifers 1800-30 510
Jonathan Palmer and John Ogden

New Zealand Temperatures, 1800-30 516
David A. Norton

Summary 521

Workshop on World Climate in 1816: a Summary and Discussion of
Results 523
Cynthia Wilson

Index 557

* The geographic sections above are not exact. For example, J. Guiot's paper, although listed
under North America, also provides substantial information on southwestern Europe and
northwestern Africa. Similarly, the paper by L. Thompson and E. Mosley-Thompson, although
listed under Southern Hemisphere, also concerns China.

Acknowledgements

The editor is grateful to his colleagues on the Organizing Committee of the international meeting
("The Year Without a Summer? Climate in 1816", Ottawa, 25-28 June 1988) from which this
volume arose: Drs. C. Wilson, A.J.W. Catchpole, T.F. Ball, R.M. Koerner, G.C. Jacoby
(Members); Mrs. Gail Rice (Secretary-Treasurer); and Mr. Kieran Shepherd (Coordinator, Poster
Presentations). I am also grateful to the following institutions for so firmly supporting the
meeting: Canadian Climate Centre; Climatic Research Unit, University of East Anglia; National
Center for Atmospheric Research (operated by the University Corporation for Atmospheric
Research under the sponsorship of the United States National Science Foundation); and the World
Meteorological Organization. I thank the directorate of the museum for its interest in and
encouragement of the project.

Joanne Dinn (Paleobiology Division) and Marie-Anne Resiga helped greatly in preparing this
book, as did Mireille Boisonneau, Arch Stewart (Canadian Museum of Nature Library) and
Daphne Sanderson (Canadian Climate Centre Library). Sharon Helman kindly redrafted several
of the figures, and with Bonnie Livingstone (Publications Division) provided strong support
during the last phases of preparing this volume.

Finally, I express my sincere thanks to Cynthia Wilson and Tim Ball for their help in organizing
the Workshop, as well as to Richard Martin for audiotaping the discussions. Cynthia Wilson
performed a particularly useful service by analyzing and summarizing the Workshop results.

Introduction

This book is the last gasp of the National Museum of Natural Sciences (now Canadian Museum
of Nature) Climatic Change in Canada Project] Because of Canada's vulnerability to climatic
change, and the lack of an integrated multidisciplinary program for studying our past climate this
project was organized. Since its beginning in 1977, a basic aim of the project has been to publish
in our Syllogeus series significant data on climatic change in Canada since the peak of the last
glaciation (about 20,000 years ago).

We began the project with a general assessment of Quaternary paleoclimatic information available
in Canada and techniques that could be used for interpreting it (Syllogeus 26, 1980); we later
broadened the number of disciplines involved and actually began gathering and interpreting the
paleoclimatic data (Syllogeus 33, 1981; 49, 1983); and then produced an annotated bibliography
on the subject (Syllogeus 51, 1984). In May 1983, the project sponsored an international meeting
"Critical Periods in the Quaternary Climatic History of Northern North America" (Syllogeus 55,
1985). It was clear from papers in Syllogeus 55 that several authors had gone well beyond the
data-gathering stage: Alan Catchpole not only tested the value of one type of proxy data (climatic
records from Hudson's Bay Company documents, including Ships' logs) against another (Marion
Parker's tree-ring records) for the Hudson Bay region, but showed that sea-ice conditions were
indicative of prevailing northerly or northwesterly winds, pumping cold arctic air over the central
and eastern parts of North America in the summer of 1816; and Cynthia Wilson took a
magnificent step forward by providing a series of six daily weather maps for early June 1816 -
actually showing the tracks of high- and low-pressure areas across central and eastern North
America.

These papers prompted me to consider convening an international meeting focusing on global
climate during 1816, "the year without a summer". What were conditions like beyond the regions
so well documented by John D. Post in The Last Subsistence Crisis in the Western World (Johns
Hopkins University Press, Baltimore, 1977) and Henry and Elizabeth Stommel in Volcano
Weather, the Story of 1816, the Year without a Summer (Seven Seas Press, Newport, 1983),
among others? Could anything useful on a global basis be added to Lamb's and Johnson's (1966,
Figure 5) excellently constructed pressure map for July 1816 extending from western Europe
across the Atlantic Ocean to central North America (Lamb, this volume)?

Accordingly, I wrote to Professor Hubert Lamb in Norwich in February 1985 and received an
encouraging, constructive reply: "Your idea of holding some sort of a conference or workshop
meeting specifically to put together the best possible reconstruction of summer 1816, or of the
whole 'year without a summer', or perhaps usefully rather more of that decade particularly aimed
at covering the period from just before the atmospheric/radiation budget disturbance caused by
the huge volcanic eruption of Tambora in 1815 till the return to the status quo ante, has intriguing
possibilities. " The proposal for this meeting was approved by the director of the museum in 1986.

The objective of the meeting was, by bringing together workers in various fields (e.g.,
volcanologists, glaciologists, climatologists, tree-ring experts, geographers, historians and
biologists) from various countries, to gain the clearest picture possible of weather and climatic
sequences in different parts of the world during 1816, or about that time (e.g., 1810-20), in an
effort to discover key factors influencing the unusual weather then. For example, how important

was the eruption of Tambora, and what other cooling influences may have been involved? How
widespread were the cold summer conditions from a global viewpoint? Did blocking play an
important part?

From the beginning, the Workshop was considered to be the heart of the meeting. The attempt
to actually plot weather and climatic data from various sources for the Tambora period on base
maps proved challenging, frustrating and exciting. Could we really shed more light on the nature
of the climatic events, their intensity and timing? Although evidence is circumstantial, it seems
that widespread cooling was underway before the eruption of Tambora. Evidently, the massive
injection of Tambora aerosols into the atmosphere in 1815 resulted in crossing a threshold to
highly anomalous weather (probably involving blocking highs and break monsoons) in many parts
of the globe. Certainly "the year without a summer" in 1816 was a regional phenomenon. In the
northern hemisphere parts of western North America, eastern Europe and Japan seem to have had
average or above-average temperatures, as opposed to the remarkable cold that characterized
much of eastern North America, western Europe, and China. Incursion of freezing arctic air
southward in one region was offset by poleward flow of tropical air in another. In the southern
hemisphere, El Nino may have diminished the cooling reflected in tree-ring records from
Argentina in 1816-17, whereas in 1817-18 the tremendous moderating influence of the Pacific
Ocean may have effectively damped any cooling recorded there (see Wilson, Workshop section,
for more details on the group's findings).

This book is intended for those who are deeply interested in: historical climate (particularly that
of the Little Ice Age) and its human impact; relationships between volcanism and climate; and
the ways paleoclimatic proxy data are gathered, treated and interpreted. The volume begins with
a general section concerning: solar influences on the trend of climate before the eruption of
Tambora; a vivid eyewitness account of the eruption; the nature of the eruption, the aerosol
produced and its course through the atmosphere - as well as a discussion of the effects of the
1783 eruption of Laki in Iceland on climate for comparative purposes and a consideration of the
effects of major volcanic eruptions following Krakatau (1883) on Canadian temperatures.
Coverage is then (loosely) geographic, first dealing with the northern hemisphere (North
America, Europe, Asia), then the southern hemisphere (South America, Antarctica, Australia and
New Zealand). Perhaps readers will gather from these contributions an inkling of the tremendous
investment in time that is presently required to distil a useful drop of paleoclimatic data from
archival and other sources.

Finally, I hope that this exercise will lead others to look more carefully at the "Tambora period"
and similar paleoclimatic problems - adding data in vast expanses of the globe where our evidence
is deficient, as well as testing and refining data given here until a more coherent picture emerges.
Information presented in this volume may also be food for ravenous paleoclimatic modellers!

C.R. Harington

General

Before Tambora: The Sun and Climate, 1790-1830

John A. Eddy1

Abstract

The unusual summer of 1816 is commonly attributed to the increase in atmospheric turbidity that
followed the eruption of Mount Tambora (Stommel and Stommel 1979). The awesome eruption
occurred, in fact, during a span of several decades of colder climate that had interrupted the
gradual global warming that followed seventeenth century extrema of the Little Ice Age (Lamb
1985). These background trends may well explain a particularly severe seasonal response in 1816
to a short-term injection of volcanic dust. The colder climate that characterized the opening
decades of the nineteenth century was quite possibly related to a coincident depression in solar
activity between about 1790 and 1830, called the "Dalton Minimum" or sometimes the "Little
Maunder Minimum" (Siscoe 1980). The probability of a solar connection is strengthened by
recent analyses of long-term changes in the level of solar activity and decadal averages of global-
surface temperature in the last 100 years (Reid and Gage 1987), as well as in the correspondence
of the Maunder Minimum in solar activity (1645-1715). A probable mechanism for solar forcing
can be found in recent spaceborne measurements of year-to-year variations in the so-called "solar
constant" (Willson et al. 1986). I plan to examine the evidence for solar and climatic anomalies
in the period from about 1790-1830 and the recent findings that provide a probable connection
between the sun and long-term climatic change.

References

Eddy, J. A. 1977. The case of the missing sunspots. Scientific American 236:80-92.

Lamb, H.H. 1985. Climate History and the Future. Princeton University Press, Princeton,
New Jersey. 884 pp.

Reid, G. and K.S. Gage. 1987. Influence of solar variability on global sea surface
temperatures. Nature 329(6135): 142-143.

Siscoe, G.L. 1980. Evidence in the auroral record for secular solar variability. Review of
Geophysics and Space Physics 18:647-658.

Stommel, H. and E. Stommel. 1979. The year without a summer. Scientific American
240:176-186.

Willson, R.C., H.S. Hudson, C. Frohlich and R.W. Brusa. 1986. Long-term downward trend
in total solar irradiance. Science 234: 1 1 14-1 1 17.

University Corporation for Atmospheric Research, Boulder, Colorado 80307, U.S.A.

Eyewitness Account of the Distant Effects of the Tarn bora Eruption of
April 1815

Michael R. Rampino1

Abstract

The following is a brief description of the effects of the eruption of Tambora volcano in 1815 on
conditions about 800 km away in eastern Java. Evidently Tambora was quite active for at least
six days prior to the cataclysmic eruption of 11 April 1815, and direct cooling was associated
with the ash cloud.

Introduction

Large explosive volcanic eruptions can have far-reaching effects on the atmosphere. The eruption
of Tambora volcano on Sumbawa Island in Indonesia in April 1815 was the largest ash eruption
in recent historic times, producing a bulk volume of about 150 km3 of pumice and ash (Stothers
1984). The loss of life and the destruction of agricultural land on Sumbawa and neighbouring
Lombok were catastrophic. In the aftermath of the Tambora eruption, in order to obtain more
information about the effects on Java and the surrounding islands, the Lieutenant Governor of
Java, Thomas Stamford Raffles, circulated a letter with three brief questions. The following
questionnaire was completed by the Resident of Surakarta in eastern Java describing local
eyewitness accounts of the effects of the Tambora eruption (catalogued in Blagden 1916). It gives
a vivid picture of the effects of the massive eruption some 800 km from the volcano. (The style,
punctuation, and spelling of the original handwritten report in the MacKenzie Collection of the
British Library has been retained throughout).

Questionnaire and Response by the Resident of Surakarta

Points of Enquiry

Circular of the Honble [T.S. Raffles] the Lieut Governor [of Java]

First, the effects of the eruption of Sumbawa in April 1815 would appear to have been first
noticed at Banjuwangie on the 1st and at Batavia on the 6th of April but the atmosphere
would appear to have been successively affected by the ashes between the 10th and 14th. On
what day and at what hour were they first noticed in different parts of your Residency -
how long - when did they continue and what was the nature of them?

At Souracarta the first explosions were heard on Wednesday the 5th of April between the hours
of 4 and 6 PM, distinct and separate sounds exceeding the number of twenty were perceived with
irregular intervals greatly resembling a military operation, but more that is denominated mortar
practice than a regular cannonade. On the successive evenings of the 6th, 7th, 8th and 9th,
occasional noises were heard which were mistaken for distant thunder. During these days the
opacity of the atmosphere, resembling former volcanic eruptions on this Island, first indicated

Earth Systems Group, Department of Applied Sciences, New York University, New York, New York 10003,
U.S.A. Also at NASA Goddard Space Flight Center, Institute for Space Studies, New York, New York 10025,
U.S.A.

the probable cause of the explosions which, by a person unaccustomed to their effects could not
be distinguished from the reports of guns or thunder etc.

On Monday the 10th, a very slight fall of dust was perceived, but alone by the most attentive
observation, and the explosions continued at intervals in the east.

On Tuesday the 1 1th the reports were more frequent and violent through the whole day: one of
the most powerful occurred in the afternoon about 2 O'Clock, this was succeeded, for nearly an
hour by a tremulous motion of the earth, distinctly indicated by the tremor of large window
frames; another comparatively violent explosion occurred late in the afternoon, but the fall of dust
was scarcely perceptible. The atmosphere appeared to be loaded with a thick vapour: the Sun was
rarely visible, and only at short intervals appearing very obscurely behind a semitransparent
substance.

The day on which the opacity of the atmosphere first commenced had not been noted accurately -
but its continuance was above twelve days, and even at the commencement of the present month
it was not entirely dissipated.

From the 5th to the 18th of the last month the Sun was not distinctly perceived, and if his rays
occasionally penetrated they appeared as observed through a thick mist. The general darkening
of the atmosphere was strikingly exhibited by such objects of which the prospect is familiar; thus
for instance at Souracarta the Mountain above continued invisible through all this period, and
even near objects were clouded or its appearance obscured by smoke -

On Wednesday the 12th the appearance of day light showed a very copious discharge of dust, this
gradually increased till 1 PM and then appeared to diminish but was still very discernible at
sunset: the following day (the 13th) it was still rarely perceptible and gradually and successively
ceased.

On the 12th a considerable darkness was occasioned by the abundance of the fall of dust: every
operation which required strong light was almost impossible within doors. The gloomy
appearance caused by the rain of dust "Udshan abu " need not be described as it was uniform in
every part of this Island to which the discharge extended. It may be remarkable that an unusual
sensation of chillings was felt during the whole of the 12th this was in great measure (tho'
probably not exclusively) occasioned by the temperature: the thermometer at 10 O'Clock AM
stood at 75 and 1/2 degrees of Fahrenheit Scale. It would appear that the subterranean
commotion, like the discharge of dust, was propogate or travelled from east to west as the
explosions were later perceived in the Western parts of the island: it is likewise highly probable
(which must however be determined by a comparison of various and accurate remarks made in
different parts of the island) that the most violent explosions were not simultaneous, but that the
combustion caused locally more powerful shocks in particular parts from Banju-wangie perhaps
to the western extremity. Something like this was remarked during the combustion of the Kloet
in 1811, when the explosions were much more violent at Batavia than at Souracarta although the
latter was much nearer to the burning Mountain. It would appear from creditable information that
effects were more sensibly felt along the Southern Shore of the Island and that the tremulous
motion of the earth was there more violent - a very uncommon rising of the water was also
perceived about the period of the most violent explosions at Harang bollong, Kadilangu etc. but
the day and hour had not been noted with sufficient accuracy for any decided inference. The
colour of the dust of the present eruption is ash grey inclining to brown it is a most impalpably

fine, divided earthy substance, if water is added it diffuses the peculiar odour of clay; it does not
acquire ductility enough to be moulded, but has been observed to improve the quality of the
common clay of the Island in the manufacture of pottery. Its chief component parts are Silecious
and Aluminous earth. It is evidently a finely divided Lava, the iron of which having by means
of gravity subsided in the vicinity of the Volcano. Scarcely any of the particles are attracted by
the magnet in this it differs from volcanic dust which was thrown from the Gunung Gunter in
1803 and, being precipitated about Batavia, possessed a blackish colour and was strongly attracted
by the magnet. The dust which was exploded by the Gunung Klut in June 1811 differs from the
present as far as can be determined without chemical analysis only by having a blueish grey
colour, and in being less finely divided; it was supposed to possess superior qualities for the
manufacturing of pottery but had not ductility enough to be moulded alone.

Have any injurious consequences resulted from within your Residency as affecting the
salubrity of the Country, or in the destruction of the Crops or Cattle respecting the latter,
state the particulars, if any and in what manner the injury may have been effected.

If the generality of the discharge of the volcanic dust is considered and the abundance of the
substance which covered the earth and of vegetation for many days, its effects on the health of
the animals were inconsiderable: instances of mortality among cattle particularly Buffaloes and
Cows in this neighbourhood during the continuance of and since the rain of dust are Solitary, and
leave it doubtful whether they must be ascribed to this or other accidental cause. In a few cases
(within my observation) death was induced suddenly: these may probably be ascribed to this
cause, but the inquiries I have made have confirmed the opinion that the health of the Cattle has
not been (in a general manner) injuriously affected. It should be kept in view in determining the
question, that previously to the rain of dust the Buffaloes in particular districts were affected by
an epidemic disease denominated Puttie by the Natives of which several died and the mortality
has in some degree continued to the present time. Neither Horses, Sheep, or Goats have been
sensibly affected.

An injury of a more serious nature threatened the crops of rice - but the forward state of
cultivation has preserved this grain in most of the neighbouring districts and such a season of
abundance as the present has not been known for many years: it has been observed by various
persons who are conversant with the cultivation of this grain, that plantations in which the rice
had nearly acquired maturity were not affected, but the dust falling upon the grain newly
transplanted in many cases destroyed the young plants. This is in some degree rendered probable
by the nature of the volcanic substance, and its effects would be more powerful towards the
period of the terminations of the rains or where a deficiency of moisture prevailed. Falling upon
the young plants and fields sparingly supplied with water it would from its clayey nature absorb
their juices and destroy them.

What was the general opinion at the time regarding the locality of the volcano?

The general opinion at this place ascribed the eruptions to the Mountain Klut of which three
previously similar "rains of ashes" were recollected by all aged inhabitants.

Conclusion

The above report documents that Tambora was quite active for at least six days prior to the
cataclysmic eruption of 1 1 April 1815. Note that the eruption was misidentified with Klut (Kelut)

volcano in Java during the ash rain. The reply gives evidence of a direct cooling associated with
the ash cloud, and such a cooling effect was observed as far away as Madras, India, where
midday temperatures fell below freezing as the cloud passed overhead (Stothers 1984). The
anomalous weather of the infamous summer of 1816 was quite likely related to the radiative
perturbation by stratospheric H0SO4 aerosols generated by the eruption. Without doubt, a similar
eruption in Indonesia today would be a regional disaster, and would create a global atmospheric
perturbation of a magnitude not seen in almost two centuries.

Acknowledgements

A grant from the American Philosophical Society supported a literature search at the British
Library for information pertaining to the aftermath of large volcanic eruptions in the nineteenth
century. The author thanks I. A. Baxter of the India Office Library, Blackfriars Road, London,
for his help, and S. Self, H. Sigurdsson, and R.B. Stothers for valuable discussions. This is a
slightly altered version of a paper published by the author in EOS 1989, p. 1559, (copyright by
the American Geophysical Union).

References

Mackenzie Collection: Private, Document 2:33, pp. 193-198, 1916. In: C.A. Blagden,
Catalogue of Manuscripts in European Languages belonging to the Library of the India
Office, Volume I: The Mackenzie Collection, Part I: the 1822 Collection and Private
Collection, p. 43. Oxford University Press, London.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science 224:1191-
1198.

The Eruption of Tarn bora in 1815: Environmental Effects and Eruption
Dynamics

Haraldur Sigurdsson1 and Steven Carey1
Abstract

New studies of deposits from the 1815 eruption of Tambora volcano provide data on eruption
dynamics, mass eruption rate and volcanic volatile emission to the atmosphere. These data form
a basis for assessment of the environmental impact of the eruption. Initial phases of activity were
two plinian explosive eruptions on 5 and 10 April with column heights of 33 and 43 km, and
mass eruption rate of 1.1x10s and 2.8xl08 kg/s respectively. The calculated column heights
therefore indicate a major injection of volcanic ash and volatile gases to the stratosphere during
the eruption. Rapid transition to pyroclastic flow generation occurred late on 10 April, when the
bulk of the material was erupted at a rate of 5.4xl08 kg/s, producing widespread co-ignimbrite
ash fall. A large component of the co-ignimbrite ash fall was produced by explosive interaction
of hot pyroclastic flows and sea water, when flows advanced into the ocean around Tambora.
Total erupted mass is estimated as 50 km3 dense-rock equivalent, or 1.4xl014 kg. Petrologic
estimates of volatile yield to the atmosphere during the eruption indicate that sulphur degassing
formed a stratospheric aerosol mass equivalent to 1.75x10" kg sulphuric acid, in agreement with
volcanic aerosol estimates based on ice-core evidence. Furthermore, volcanic degassing of 10"
kg HCf and 7.4xl010 kg HF occurred, but the fate of these species in the atmosphere is
unknown. Climatological data indicate a short-term northern hemisphere surface temperature
decrease of 0.7 °C following the eruption, and this climatic response agrees with the empirical
relationship observed between sulphuric acid volcanic aerosol mass and temperature decline
observed after several major explosive volcanic events. It is likely, however, that the observed
surface temperature decline is not solely due to the Tambora event, as a cooling trend was
already in progress prior to the eruption.

Introduction

The 1815 Tambora eruption on the island of Sumbawa in Indonesia exceeded in magnitude any
other volcanic eruption in historical times, producing over 50 km3 of magma. As a measure of
the uniqueness of this great natural disaster, it is remarkable that we have to search some 20,000
years back in the geological record to find an explosive eruption of greater magnitude: the
Shikotsu eruption in Japan (Katsui 1959). The volume of material erupted from Tambora is an
order of magnitude greater than that discharged in the celebrated Krakatau eruption of 1883, and
two orders of magnitude greater than in the 1980 Mount St. Helens eruption. Locally, 92,000
people died on Sumbawa and adjacent islands, either directly from effects of the eruption or from
the ensuing famine and epidemic. In addition to its significance as a geological process, the
eruption had unprecedented impact on the Earth's stratosphere. The eruption injected enormous
quantities of sulphur, chlorine and fluorine gases into the stratosphere, leading to a variety of
global atmospheric phenomena, and was probably responsible for the marked climatic
deterioration of 1816. Thus, although the eruption is of great importance to the study of
volcanology, its greatest scientific significance probably relates to the environmental effects, i.e.,

Graduate School of Oceanography, University of Rhode Island, Kingston, Rhode Island 02881, U.S.A.

its impact on the chemistry of the atmosphere and on climate (see also Vupputuri, this volume).
With the growing realization of connections hetween the biosphere, atmosphere and geosphere
and the recognition of global environmental and climatic change brought about by human activity,
the detailed study of the effects of Nature's own large-scale experiments such as the Tambora
eruption can greatly aid in our understanding of short- and long-term changes in the global
environment.

In this paper we summarize our findings based on a new study of the Tambora deposits,
involving two expeditions to the volcano in 1986 and 1988, which provide fresh data on the
eruption dynamics and erupted mass. In addition, our recent petrologic study of the sulphur,
chlorine and fluorine yield of the eruption to the atmosphere gives quantitative estimates of
degassing and provides a framework for modelling the environmental impact of this great
volcanic pollution event.

Chronology of the 1815 Eruption

Before 1815, Tambora volcano was conical in form, possibly with two peaks, and the highest
mountain in the Sunda Islands. When sailing east from Java, Tambora appeared equally
prominent on the horizon as the 3,726-m high Rinjani volcano on Lombok, and Zollinger (1855)
estimates that the volcano may have been over 4,000 m before the eruption. His estimates are
based on discussions with people in Sumbawa, who maintained that the volcano had lost at least
one third of its height. The maximum height of the caldera rim after the eruption is 2,850 m.

Contemporary local sources about the 1815 Tambora eruption are mainly newspapers and
government accounts - particularly the Asiatic Journal. These accounts are especially useful for
establishing the timing of various eruptive phases, and the extent and nature of their effects. This
section summarizes important eyewitness observations that are relevant to interpreting the
pyroclastic deposits studied in the field (Rampino, this volume).

More than three years before the great eruption, a thick cloud had formed over the peak, which
not even the strongest winds could dissipate (Zollinger 1855). It gradually grew darker and
larger, and extended farther down the volcano's flanks. Explosions were heard from the volcano
during this time; first only a few and weak, but gradually they became more frequent and louder.
People living around the volcano sent delegates to the government authorities in Bima on
Sumbawa requesting an investigation of these phenomena. The authorities sent a man by the name
of Israel, whose brother was still alive at the time of Zollinger's visit. Israel reached the Tambora
region on the evening of 9 April, the day before the climax of the eruption, and was killed during
the activity the following day.

On the evening of 5 April, the first major eruption began and was heard widely in the Indonesian
region. The explosions heard in Java resembled cannonfire and soldiers in Yogyakarta (central
Java) combed the land and seas for invaders (Figure 1). Ash fell "like fine snow" in Banjuwangi
in eastern Java, accumulating up to one-half inch (1.3 cm) thickness. Minor ash fall also occurred
at Besuki in east Java. At Solo (central-eastern Java, 800 km from Tambora) sounds of explosions
commenced on the evening of 5 April. The naval vessel Benares was in Macassar on 5 April,
about 350 km NNE of Tambora (Figure 1). Loud explosions were heard from the south, which
continued the entire afternoon. At sunset the explosions grew louder and seemed closer. Listeners
suspected that a naval battle was taking place nearby and sent troops to search the region.

Figure I: The Indonesian region, showing areas where sounds were heard from the 1815 Tambora
eruption.

Ash fell in east Java during the morning of 6 April, but only a trace descended on western Java.
The sky gradually cleared during the day, but the air was hot and the atmosphere unusually still.
Between 6 and 8 a.m. on 6 April, loud noises were heard at Ternate, where the ship Teignmouth
lay at anchor some 1,400 km northeast of Tambora.

After four days of minor activity, the volcano became very active again on 10 April. Witnessing
the activity from Sanggar on the eastern slopes of Tambora, the Rajah described the second and
larger major eruption. At about 7 p.m. three columns of fire rose high from Tambora's crater,
uniting in a single firestorm over the volcano. Moments later the entire mountain was a sea of
glowing flows, which spread in all directions. Large quantities of ash and stones fell on Sanggar
(Figure 2), "up to two fists in size", but most were no larger than a nut. Between 9 and 10 p.m.
the ash fall increased, and a strong "whirlwind" descended carrying off houses in Sanggar and
nearby villages. In the part of Sanggar nearest to Tambora, the largest trees were uprooted by
the windstorm, and carried off with houses, people and livestock. These descriptions are
consistent with the passage of a pyroclastic surge through the village. Sea level rose suddenly
12 feet (3.7 m).

At Bima, 80 km east of the volcano (Figure 2), the explosions sounded like heavy mortar fire
during the night of 10 to 11 April. The town was in complete darkness from the ash cloud
overhead from 7 a.m. on 1 1 April to 14 April. Ash fall was so heavy, that roofs of most houses
collapsed. The air was completely still and there was no wind at sea, but nevertheless the waves
were very high and flooded the coast and into the town. All boats were torn from their moorings
and tossed ashore.

20 km

Scale

Flores Sea

Nguvu

Figure 2: Sumbawa Island, showing Tambora volcano and other places referred to in the text.

Beginning on 10 April, thunderous noises were heard in many parts of Java, which were much
louder than on 5 April, especially east of Cirebon (western Java, 1,050 km from Tambora). At
Banjuwangi in east Java (400 km distant) the evening noises were very loud and shook the earth.
The sounds became somewhat weaker toward morning the next day, but continued until 14 April.
At Sumenep (Madura Island, 470 km distant) the noises were like rapid cannonfire. The sky was
completely obscured by ash and, in some districts such as Solo and Rembang (central Java,
Figure 1), earth vibrations were felt.

During the night of 10 to 11 April the Benares reported from Macassar that explosions began
again and grew in frequency the next morning, shaking both houses and ships. Lightning flashes
were common and the sky was very dark, especially to the south and southwest. The sea rose

rom five to seven feet (1.6 to 2.2 m) above normal in Besuki, eastern Java, on the night of 10

pril.

On 1 1 April the continuing activity was so severe that houses shook in eastern districts of Java.
The coast of Bali was totally invisible from Banjuwangi in eastern Java, where candles were lit
at 1 p.m. By 4 p.m. it was pitch-dark and remained dark until 2 p.m. the next day. In Sumenep
(Madura) the light was so faint that candles had to be lit before 4 p.m. The following night was
indescribably dark. At about 7 p.m. a tidal wave struck Sumenep Bay raising sea level about four
feet (1.2 m) for several minutes. Major ash fall also began in Besuki, eastern Java, on 11 April,
with darkness extending from 4 p.m. on 1 1 April until 2 p.m. on 12 April. Explosions were also
heard on 11 April at Ambon (Asiatic Journal, February 1816, p. 116).

A boat sailing from Timor in the east noted that the sky became very dark as they approached
Tambora on 1 1 April. When they were off Tambora, the base of the volcano was engulfed in
flames and the peak was shrouded in a dark cloud, with fires and flames shooting out. They went
ashore for water in Sumbawa and found that all boats had been cast ashore by tidal waves. They
came across a large number of corpses. As they sailed from Sumbawa, they encountered large
rafts of pumice, which formed thick layers on the ocean hindering their passage. Some pumice
rafts were so thick that they resembled sandbanks or low cliffs. They were caught in a pumice
raft over two feet (0.6 m) thick the entire night of 12 April. The vessel Dispatch heard explosions
on the night of 11 April, when about 7° east of Bima (about 750 km). Rafts of pumice and
timber were so thick along the coast of Flores, that the ship had great difficulty in making way.

Effects of the eruption were noted as far west as Sumatra. On the morning of 11 April, loud
noises were heard at Bengkulu on the south coast of Sumatra about 1,800 km west of Tambora,
and as far as Terumon in western Sumatra some 2,600 km WNW of Tambora. Explosions were
also heard on Bangka Island (1,500 km) off the northeastern coast of Sumatra. People from the
interior of Sumatra reported that the leaves of trees and crops were covered with a layer of very
fine ash {Asiatic Journal, June 1816, p. 600 and August 1816, p. 164).

On 12 April only very faint daylight was visible in eastern Java, and objects were barely visible
at a distance of 100 paces in Solo. Some light returned in Banjuwangi about 2 p.m., but the sun
was not visible until 14 April. It was unusually cold during this period. Ashfall in Banjuwangi
was nine inches (of which eight inches (20.3 cm) had accumulated by 12 April), two inches
(5 cm) in Sumenep and somewhat less in Gresik. West of Samarang (central Java) the daylight
was little affected.

At 8 a.m. on 12 April it was dark on the Benares in Macassar, and by 10 a.m. it was so dark
that nearby ships could not be seen. By 1 1 a.m. the sky was completely dark, except for a small
clearing in the east. The ash fell as heavily as snow, and the sea and air were still. By 12 noon
the faint light in the east had vanished and it was so dark that a hand held in front of the face
could not be seen. Ash fell all night and was so fine that it penetrated all parts of the ship below
decks. By 13 April the intensity of the eruption had decreased but its effects were still
widespread. At 6 a.m. it was still totally dark in Macassar but faint light returned at 7:30 a.m.,
and by 8 a.m. one could discern objects. Sounds of the explosion ceased the following day at
Banjuwangi but ash fall continued in Macassar, accompanied by a calm and great heat until
15 April.

Not until 17 April did the ash fall cease, and heavy rains spread over the region. In Banjuwangi,
many houses had collapsed under the weight of the ash, and fever and epidemics had broken out
in several regions affected by the ash fall. On Java the damage to livestock and agriculture was
most severe in the eastern district around Banjuwangi, where the destruction of crops and grazing
areas was so extensive that many horses and cattle died of hunger.

The devastating effects of the eruption on the local population were first realized when ships
reached ports on Sumbawa. Benares reached the coast of Sumbawa on 18 April and was trapped
in large pumice rafts on the sea. The rafts were so large, that they at first took them for
sandbanks or new islands: they were often over a nautical mile in length, and had varied surface
features. Large numbers of carbonized and splintered trees were trapped in the rafts. Benares
dropped anchor at Bima on 19 April, where ash fall was 3% inches (9.5 cm) thick. The harbour

had changed, and they found eight fathoms (14.9 m) where the depth had been six fathoms
(11.2 m) before the eruption.

Some pumice rafts were up to three miles (4.8 km) long, and were still troublesome to navigation
between Moyo and Sanggar three years after the eruption. Pumice rafts from the volcano drifted
widely over the southern seas in the following months. Between 1 and 3 October 1815, the ship
Fairlie, in the Indian Ocean on passage to Calcutta, sailed for two days through extensive pumice
rafts, about 3,600 km west of Tambora (Asiatic Journal, August 1816, p. 161). These rafts
travelled at a rate of 0.2 m/s from Tambora and were most likely transported in the South
Equatorial Current, driven by the southeast trade winds. Ash fall from the eruption also reached
Brunei in Borneo, where the phenomenon so impressed the local people, that they subsequently
counted the years from "the great fall of ashes" (Reclus 1871).

On 22 April, the Dispatch arrived in Bima. It had first dropped anchor near Sanggar, where the
Rajah had told them that all the land was now a desert and all crops and fruits were destroyed.
Sanggar Bay was covered with pumice rafts including large trees and remains of houses carried
out to sea by the eruption. The volcano was still covered in dense clouds of ash and steam.
Smoke emanated in many places from hot flows of ash on the lower flanks, which had also
entered the sea.

The British Governor of Java sent Lieutenant Owen Philipps to Sumbawa to study the event and
its effects on the people. On the way from Bima to Dompu (Figure 2), Philipps observed a large
number of corpses along the road. Villages were abandoned and houses had generally collapsed
under the weight of the ashfall. The few survivors wandered about in search of food. The
population had been affected by severe diarrhoea, which had caused many deaths. The people
blamed this on their drinking water, which was contaminated with the volcanic ash. Horses and
other livestock were also killed in large numbers by this disease. The Rajah of Sanggar met with
Philipps in Dompu. The misery of his people was much worse than in Dompu and even one of
the Rajah's daughters had died of hunger. Coconuts were the only food supply of the ruined
village, where starvation was severe. Philipps gave him some rice, for which the Rajah gave
thanks with tears in his eyes.

Zollinger (1855) describes the misery of the remaining population. Many continued to wander
in search of food and willingly sold themselves as slaves, sometimes for a few pounds of rice.
His studies indicate that about 10,100 people died in Sumbawa directly by the effects of the
eruption, most likely in pyroclastic flows and surges (Table 1). Contemporary estimates of
number of fatalities in several villages vary. Thus, for example, Tobias claims there were 10,000
deaths in Tambora village alone, whereas Philipps claims 12,000 victims in this village. In
addition, 37,825 died by starvation and 36,275 migrated from Sumbawa. Zollinger estimates that
at least 10,000 died in Lombok from starvation and disease, but the loss there was much more
severe according to Van der Broeck (1834), who states that the population of Lombok was
reduced from 200,000 to 20,000 by the effects of the eruption. Zollinger claims his numbers are
all minimum estimates. Junghuhn (1850) estimates that the fatalities on Sumbawa were 12,000
and that 44,000 died on Lombok, but his estimate does not include the starvation victims on
Sumbawa. The most-quoted fatality figures of the eruption are those of Petroeschevsky (1949),
who estimates that the total number of victims was 92,000 - 48,000 on Sumbawa and 44,000 on
Lombok, or 35 and 22.5% of the estimated total population of these islands, respectively.

Table 1: Fate of the Human Population in Sumbawa.1

Village

Eruption Victims

Death by Starvation

Refugees

Pekat

2,000

Tambora

6,000

Sanggar

1,100

825

275

Dompu

1,000

4,000

3,000

Sumbawa

18,000

Bima

15,000

Totals:

10,100

37,825

36,275

1 After Zollinger (1855).

The only village near Tambora that remained undamaged was Tempo, with 40 inhabitants. Of
the total population of 12,000 of Tambora and Pekat, only five or six survived. All trees and
vegetation north and west of the volcano were completely destroyed, with the exception of a high
point near the village of Tambora. Zollinger remarked on the long-term effects of the eruption
on Sumbawa's climate and vegetation. Soil became very dry, rainfall decreased, and all
vegetation suffered a severe setback, and would take an estimated several hundred years to
recover fully.

Pyroclastic Deposits from the 1815 Eruption

As a consequence of the eruption, the upper part of the volcano collapsed to form a 6-km
diameter, 1 ,200-m deep caldera with a total volume of about 28 km3, and Tambora lost about
1,200 to 1,400 m of its height, corresponding to about 6 km3 or a total of 34 km3. The void
formed by the caldera collapse represents in part rock formations ejected from the volcano, and
in part the subsidence of the volcano's edifice into the underlying magma chamber. The former
can be evaluated from proportion of lithics in the fall deposits, which is about 5.5 wt.%
(Sigurdsson and Carey in press, Table 2) or less than 4 km3 of rock. Ejection of solid rock can
consequently account for one-tenth of the caldera volume. The total ejected mass of magma is
1.3xl014 kg, less the lithics, corresponding to about 50 km3 of magma withdrawn from the
reservoir - substantially larger than the observed caldera volume. Subsidence into the emptying
magma reservoir is regarded as the dominant mechanism of caldera formation.

The deposits laid down outside the caldera during the eruption reflect two major processes:
(1) early explosive activity (plinian and phreatomagmatic) producing high eruption columns and
four tephra or ash-fall deposits; and (2) subsequent ignimbrite phase activity during collapse of
the eruption column, producing at least seven pyroclastic flows and surge deposits, with
associated large-volume co-ignimbrite ash falls.

Table 2: Tambora 1815; Composition of Glass Inclusions in Plagioclase Phenocrysts.1

Si02

57.37 (1.28)

57.01 (.51)

56.37 (.29)

56.88 (1.24)

56.58 (1.13)

Ti02

0.6 ( .07)

0.56 (.06)

0.73 (.01)

0.60 (

.10)

0.72 ( .31)

ALO3

19.66 ( .41)

19.58 (.22)

19.43 (.06)

19.88 (

.45)

20.17 ( .21)

FeO

4.56 ( .39)

4.47 (.20)

4.66 (.07)

4.73 (

.48)

5.23 ( .94)

MnO

0 27 ( 05)

0 28 ( 08)

0 25 ( 01)

0.19 (

.08)

0 24 ( 05)

MgO

1.33 ( .17)

1.09 (.07)

1.18 (.03)

1.37 (

.15)

1.75 ( .44)

CeO

2.71 ( .35)

3.09 (.08)

2.87 (.01)

2.85 (

.25)

2.50 ( .25)

Na.0

6.19 ( .25)

5.5 (.50)

5.89

6.44

3.15 ( .54)2

5.09 ( .47)

5.59 (.27)

5.69 (.09)

5.35 (

.86)

6 00 ( 48}

0.06 ( 0)

0.04 (.02)

0.31 (.13)

0.36 (

.15)

Total

97.68

97.21

97.38

98.65

96.34

Number of

Inclusions

Number of

Analyses

Volatiles by

Difference

2.32

2.79

2.62

1.35

Water by

Difference3

1.95

2.42

2.25

0.98

Sulphur (ppm)

512±62

589 ±94

613± 157

381 ± 44

Chlorine (ppm)

1,747 ±337

2,057 ±732

2,375±532

2,817±1253

2,106±163

Fluorine (ppm)

1,185± 87

1 - glass inclusions in plagioclase from plinian fall layer F-2, sample TB-42; 2 - glass inclusions in plagioclase from
lower part (0 to 5 cm) of plinian fall F-4, sample TB-86; 3 - glass inclusions in plagioclase from upper part (10
to 15 cm) of plinian fall F-4, sample TB-88; 4 - glass inclusions in plagioclase from co-ignimbrite fall deposit F-5,
sample TB-136; 5 - glass inclusions in plagioclase from tephra fall, sample T58-A (Devine el al. 1984).

Not corrected for sodium loss during microprobe analysis. All other values are corrected.

Water by difference is calculated as volatiles by difference minus S, CI and F
(0.37 wt.%).

The four initial explosive events produced widespread tephra fall deposits, which can be traced
at least to Lombok, 150 km west of the caldera (Sigurdsson and Carey in press). The basal F-l
ash fall is the product of phreatomagmatic explosions, resulting from interaction of magma with
the hydrothermal system of the volcano (Figure 3). Historical evidence (Petroeschevsky 1949)
indicates that the volcano was mildly active in the period 1812-15, when "rumblings and dense

clouds" were noted. The F-l ash fall probably originated during this early activity, as magma was
making its way from a deep reservoir toward the surface and periodically erupting in small
outbursts. The total volume of tephra erupted during the phase of activity was about 0. 1 km3.
Evidence from our excavations in the ruins of the ancient Tambora village, 2 km east of Tambora
Coffee Estate, indicates that the F-l event took place long before the subsequent activity. The F-l
layer is absent from the village, indicating its complete erosion before the 5 April eruption.

Stratigraphy of 1815 Tambora Deposits

April 10-11

April 10

April 5-10

April 5, 1815

Pre-April 5, 1815

Figure 3: Stratigraphy of the 1815 pyroclastic deposits in a typical section at Gambah on the
northwestern slopes of Tambora volcano, 25 km from the caldera. Dates on the right of the
stratigraphic column indicate the timing of successive eruptive phases, based on historical
reports.

The F-2 pumice fall layer marks the first major explosive eruption during 1815. The distribution,
lithology and grain-size of this deposit indicate typical plinian activity. About 1.2 km3 of material
was ejected. We correlate this plinian eruption with the explosion of 5 April 1815 that was heard
in Jakarta (1,250 km away) and Ternate (1,400 km away) and caused ash fall as far as Besoeki
in East Java (Raffles 1835).

r 60 cm

-50

-40

-30

-20

- 10

PF-I Pyroclastic flow
(1-4 meters )

S-l Pyroclastic surge

F-4 Plinian Pumice Fall

F-3 Phreatomagatic Ash Fall

F-2 Plinian Pumice Fall

F-l Phreatomagmatic Ash Fall

After the F-2 plinian event, Tambora lapsed into a state of low-level activity from 5 to 10 April.
During this period, several smaller explosions produced tephra fall, which forms layer F-3
(Figure 3). The deposit is highly fragmented, like the first layer of the eruption (F-l), but the
evidence of the phreatomagmatic activity is not as compelling.

A second major plinian eruption produced the F-4 pumice fall layer. Grading of this deposit
shows a rapid rise of the eruption column during the first third of the eruption, followed by a
slow decline. The F-4 layer is much thicker and coarser than the earlier F-2 plinian fall, although
similar in lithology. Despite its high intensity, this phase of the eruption ejected only a moderate
amount of material (3 km3 of tephra). The F-4 fall deposit is clearly from the beginning of the
10 April paroxysmal event.

The Rajah of Sanggar reported an intensification of activity at about 7 p.m. on 10 April, followed
by a rain of pumice on Sanggar, east of the volcano, at approximately 8 p.m. Tephra fall
continued until about 10 p.m. when the village experienced winds that uprooted trees and
buildings. The whole volcano appeared like a flowing mass of "liquid fire". This event is marked
clearly in the volcanic stratigraphy everywhere on Sanggar Peninsula by the abrupt transition
from F-4 plinian pumice fall to the overlying charcoal-bearing surge and pyroclastic flows (Figure
3). The change in the eruption mechanism may have been primarily due to continued vent erosion
during F-4 plinian activity, leading to eruption column collapse, with resulting pyroclastic flows
and surges. No significant time break may have occurred during the transition.

In distal localities the F-4 plinian fall is overlain by a 12- to 25-cm thick greyish-brown, poorly-
sorted, silty-sandy ash (F-5). Unlike the other fall deposits from the 1815 eruption that show
systematic thinning with distance from source, the F-5 ash fall retains a constant thickness to a
remarkable degree - in fact thickening appreciably to the west, away from the volcano.
Consequently the F-5 layer represents an increasing proportion of the total fall with distance from
source, increasing from about 25% of the total fall deposit thickness at 40 km, to about 80%
beyond 90 km (Figure 4).

The F-5 layer does not correspond to any fall deposits in the proximal area, but is
stratigraphically equivalent to the surges and pyroclastic flows. We therefore consider the F-5
deposit formed primarily from ash and pumice fallout from an eruption column generated during
the surge and pyroclastic flow phase, i.e., a co-ignimbrite and co-surge ash fall deposit. The co-
ignimbrite ash fall was not only generated by glass elutriation from the convecting eruption
columns and flows, but also by wholesale depletion of the fine fraction of crystals and glass alike
from the column and flows. We attribute this depletion to explosive interaction between
pyroclastic flows and the sea along the coast of the Sanggar Peninsula (Sigurdsson and Carey in
press), based on comparative grain-size studies of inland and coastal pyroclastic-flow deposits.
Our model proposes the creation of large secondary eruption columns around the peninsula of the
volcano, where high-temperature pyroclastic flows were discharged into, and reacted explosively
with seawater. The secondary plumes consisted mostly of fine-grained (<200 micron) ash and
steam.

Total Erupted Mass

It is generally recognized that the Tambora eruption involved an exceptionally large volume of
magma, although quantitative estimates have varied greatly. Thus, Zollinger (1855) estimated the
ash fall volume at > 1,000 km3, Junghun (1850) 318 km3, Verbeek (1885) 150 km3, Sapper
(1917) 140 km3, Pannekoek van Rheden (1918) 30 km3, and Petroeschevsky (1949) estimated

total ash fall of 100 km3 on the basis of observed thicknesses. A reassessment of the ash fall
volume by Stothers (1984) led to an estimate of 150 km3, and Self et al. (1984) estimate 175 km3.
New estimates can now be made on basis of our recent field work, Sigurdsson and Carey (in
press).

30 1

| 20

2 10

40 80 120

Distance from Source (km)

160

1.0
0.8

3 0.6 i

S 0.4

"3 0.2

0.0

40 80 120
Distance from Source (km)

160

0.7

.2 06
I 0.5

M 0.4-
5P

0.3
0.2
0.1

40 80 120
Distance from Source (km)

160

Figure 4: Characteristics of the F-5 co-ignimbrite tephra fall deposit as a function of distance from
source, showing: (a) variation in thickness in cm; (b) thickness of F-5 co-ignimbrite ash fall
as a fraction of total ash fall thickness; (c) crystal/glass ratio of the co-ignimbrite ash fall.
Horizontal line in (c) is the crystal/glass ratio in the erupted magma, as determined in
artificially-crushed pumices from the pyroclastic flows.

As shown above, the products of the eruption form a multi-layer deposit, reflecting several
processes in action. The four early fall deposits produced during activity from 5 to 10 April (F-l
to F-4), have a total volume of 4.6 km3, corresponding to 1.8 km3 of dense rock, or about
4.3xl012 kg of magma. While this represents an eruption larger than the 1980 Mount St. Helens
event, and comparable in volume to the 1982 El Chich6n event, these early April falls from
Tambora represent only 5% of the total erupted mass in 1815. The co-ignimbrite ash-fall layer

F-5 is the dominant part of the deposit. Volume of the total ash fall can be estimated from
contemporary accounts of ash fall and thickness, and deep-sea core evidence (Neeb 1943). Using
the isopach map compiled by Self et al. (1984) and shown in Figure 7, we estimate that 90 km3
of tephra was deposited within the 1 micron isopach, corresponding to 22 km3 dense-rock
equivalent of distal fall (5.5xl013 kg). The F-5 fall must represent about 92% of this volume, as
the combined volume of the earlier F-l to F-4 fall layers is only 1.8 km3 DRE. An estimate of
density of the deposit is required in order to assess the erupted mass. During the eruption, ash
fell on decks of the ship Benares near Macassar in Sulawesi. A pint of the ash was reported to
weigh 1214 oz, corresponding to a deposit density of 611 kg/m3 of the fresh-fallen ash (Asiatic
Journal 2, 1816, p. 166). A minimum mass of 5.8xl013 kg is therefore represented by the fall
deposit.

Studies of the volcano and its deposits indicate that a large mass of pyroclastic flows entered the
ocean during the eruption (Sigurdsson and Carey in press). We estimate a pyroclastic flow deposit
volume of about 30 km3, equivalent to 8.2xl013 kg. Thus, the total erupted mass is of the order
1.4xl014 kg of magma. No historical eruptions have produced as large a mass of magma as
Tambora, which emitted more than twice the mass of the nearest large-magnitude event, i.e., the
1783 Laki eruption in Iceland.

Mass Eruption Rate and Column Height

Pumice and lithic isopleth maps of the F-2 and F-4 layers are presented by Sigurdsson and Carey
(in press). The area encompassed by a specific isopleth is considerably larger for the F-4 layer,
demonstrating the greater intensity and dispersal of that plinian event. The distribution of the two
Tambora plinian layers is compared with several other well -documented plinian fall deposits
(Figure 5). Our new isopleth data indicate that the two Tambora plinian fall deposits had greater
dispersal than any plinian eruption in historic times. Despite the fact that their dispersal compares
with some of the largest known plinian fall deposits in the geological record, the thicknesses and
thus volumes of the two Tambora fall deposits are relatively small (Figure 5). The great dispersal
of clasts during the two plinian eruptions of Tambora is noteworthy and has important
implications for existing models of the 1815 eruption.

The dispersal characteristics of the pumice fall preserve information about the dynamics of the
eruption column and the atmospherically-dispersed plume. Thus, the geometry of lithic isopleths
can be used to determine the maximum eruption-column height and average wind speed for a
specific fall layer (Carey and Sparks 1986). The half-width of an isopleth measured perpendicular
to the main dispersal axis is primarily a function of the eruption column height, whereas the
maximum downwind range along the axis is controlled by both column height and average wind
speed. Data from the 3.2-cm diameter lithic isopleths of the F-2 and F-4 layers indicate eruption-
column heights of 33 and 43 km, respectively (Figure 6). This places the F-2 column higher than
the maximum height achieved by the 79 A.D. plinian eruption of Vesuvius (Carey and Sigurdsson
1987), and the F-4 column is slightly higher than the great 1956 Bezyminanny eruption
(Gorshkov 1959). The F-4 plinian phase is thus the most energetic plinian activity ever recorded
in historic times, and is exceeded in intensity by only one eruption in the geological record - the
"ultraplinian" Taupo pumice fall in New Zealand (Walker 1980).

lo-i — ' — — 1 I i

1 10 100 1000 0 20 40 60 80

Lithic Diameter (mm) Distance from Source (km)

Figure 5: Comparison of 1815 Tambora fall deposits with characteristics of deposits from other major
volcanic eruptions, (a) Plot of lithic isopleth area versus lithic diameter for the F-2 (5 April)
and F-4 (10 April) plinian fall deposits compared with the plinian falls from the eruptions of
Vesuvius, Italy (79 AD), Osumi, Japan, and Taupo and Waimihia, New Zealand, (b) Plot of
thickness versus distance from source for the F-2 and F-4 Tambora plinian fall layers
compared to other well-known plinian deposits, as in Figure 5 (a). Note that despite the fact
that Tambora layers are very widely dispersed, they are substantially thinner than other major
plinian fall deposits.

0 5 10 15 20 25 30

MAXIMUM DOWNWIND RANGE (km)

Figure 6: Plot of isopleth half-width versus maximum downwind range of the 3.2-cm diameter lithics
isopleths for the F-2 and F-4 plinian Tambora fall deposits, compared with other well-known
plinian falls. Diagonal lines are wind-velocity contours in m/second, and horizontal lines are
maximum eruption-column heights in kilometres. Note the 43 km high F-4 plinian column
from 10 April 1815 above Tambora volcano, and the 33 km high column from the F-2
eruption on 5 April.

The estimates of eruption-column height can be used to calculate the eruption rate by using
relations for a tropical atmosphere (Sparks 1986). Our calculations indicate a rate of 1 . lxlO8 kg/s
for the F-2 phase and 2.8xl08 kg/s for the more energetic F-4 event. With these values it is
possible to estimate the duration of the events by simply dividing the total mass of tephra in each
layer by the perspective rate of magma discharge. Assuming that the maximum magma-discharge
rate was active throughout the plinian eruptions, each layer would have been ejected in 2.8 hours.

Duration of the co-ignimbrite fall was about three days, judging from the historical reports, e.g.,
from Madura Island, 500 km WNW of the volcano (Figure 7). This is the period of the
sedimentation of tephra from the atmosphere and thus represents the maximum duration of the
eruption which began on 10 April. In order to accommodate the total ignimbrite and co-
ignimbrite mass in this period (1.4xl014 kg), we infer a minimum ignimbrite mass-eruption rate
of the order of 5.4xl08 kg/s, or about three times the peak rate during the eruption of the
preceding F-4 plinian fall. This is about half the rate of the highest intensity event known: the
Taupo eruption in 130 AD, with an eruption rate of l.lxlO9 kg/s (Walker 1980).

Figure 7: Isopach map of the total ash fall from the 1815 Tambora eruption, based on contemporary
reports of ash fall and evidence from bottom samples collected during the Snellius Expedition.

Volatile Emission from the Tambora Eruption

Recent studies have shown that quantitative estimates can be made by petrological methods of the
mass and type of volatiles (e.g., sulphur, chlorine and fluorine) released during volcanic
eruptions. The potential of trapped glass inclusions as recorders of pre-eruption volatile content
of magmas was first recognized by Anderson (1974), who applied this method in estimating the
volcanic volatile contribution to the sulphur and chlorine budget of the oceans. The method was
also applied in the 1976 St. Augustine eruption by Johnston (1980), who demonstrated the
potentially great contribution of volcanic eruptions to the chlorine budget of the stratosphere.
These studies paved the way for the petrologic estimates of volcanic degassing during earlier and
prehistoric eruptions (Sigurdsson 1982; Devine et al. 1984; Palais and Sigurdsson 1989). When
compared with other determinations of volatile emission based on ice-core acidity and
atmospheric observations, the petrologic estimates yield similar results for the same eruptions
(Sigurdsson et al. 1985).

In the first petrologic study of volcanic volatiles from the 1815 Tambora eruption, Devine et al.
(1984) found that seven glass inclusions in feldspar phenocrysts from a single pumice sample
contained on the average 380 ppm sulphur, 2,100 ppm chlorine, 1,190 ppm fluorine. We have
analyzed glass inclusions in plagioclase phenocrysts and matrix glasses in five tephra samples
from the 1815 eruption, representing all major deposits produced during the event. Our results
(Tables 2, 3) show that the eruption tapped a homogenous body of trachyandesite magma, with
no systematic chemical gradients. We find that the average pre-eruption concentration of volatiles
is 570 ppm sulphur, 2,220 ppm chlorine and 1,190 ppm fluorine, whereas the degassed matrix
glass has on the average 266 ppm sulphur, 1,486 ppm chlorine and 680 ppm fluorine. These
results indicate, that about 53% of the pre-eruption sulphur content of the magma was lost to the
atmosphere during the eruption, accompanied by loss of 33% of the chlorine and 43% of the
fluorine. In addition, the results indicate a pre-eruption water content of about 2 to 2.4 wt.% in
the magma.

Table 3: Matrix Glass Composition of Tambora 1815 Tephra (parts per million).1

Sample Numbers

TB-42 TB-86 TB-88 TB-136 TB-87 T58-A

Sulphur 363 ±57 126 ±17 241 ±67 196 ±48 362±33 309 ±7

Chlorine 1,523±174 1,460±142 1,621 ±22 1,476±69 1,627± 139 1,211 ±50

Fluorine - - - - - 679 ±69

Errors are one standard deviation of the average.

With a known total erupted mass of magma of 1.4xl014 kg, the minimum mass of volatiles
emitted to the atmosphere can be estimated from the difference in volatile concentration between
glass inclusions and matrix glasses. These calculations show that about 4.3xl010 kg of sulphur
were released to the atmosphere, 1x10" kg chlorine, and 7xl010 kg fluorine. These improved
estimates are somewhat lower than the preliminary values of Devine et al. (1984) for Tambora
volatile degassing, but still place Tambora as the pre-eminent volcanic pollution event in historic
time, with a total mass of 2.1x10" kg of sulphur, chlorine and fluorine released to the
atmosphere. Further studies of the poorly-constrained volume of the distal ash fall will probably
lead to an increase in these estimates. In addition, we infer that about 2.8xl012 kg of magmatic
H20 was introduced into the atmosphere during the eruption, or equivalent to more than doubling
the stratospheric water-vapour content. Further addition of large quantities of meteoric water
vapour to the stratosphere resulted from the large-scale convective flow of humid tropospheric
air, entrained in the ascending eruption column.

No measurements have been made of carbon dioxide levels in the Tambora products, but some
inferences can be made of C02 output from the eruption. Magma of the type erupted from
Tambora in 1815 is likely to have C02 levels of the order 500 ppm, judging from the solubility
data of Stolper and Holloway (in press). Degassing of magma of this type would then yield about
1014 g C02 to the atmosphere during the 1815 eruption. Thus, the carbon dioxide output from
Tambora would be roughly equivalent to the annual output from the Earth's mantle, and only
about 1 % of the current annual anthropogenic output of C02.

Sulphur Aerosol

Sulphur output from Tambora during the three-day period in 1815 was more than double the
current annual total sulphur output of volcanoes, which has been estimated as 0.9 to 1.2X1010
kg/yr (Berresheim and Jaeschke 1983; Stoiber et al. 1987). In comparison, the annual global
anthropogenic emission rate of sulphur dioxide is estimated as 1.3x10" kg (Bach 1976). The fates
and atmospheric effects of anthropogenic and volcano-derived sulphur aerosols are, however,
quite different. The anthropogenic emission, caused by burning of fossil fuels, is mostly confined
to the troposphere, where its residence time is short. In contrast, highly energetic explosive
volcanic eruptions transport sulphur and other volatile species rapidly to the upper troposphere
and lower stratosphere. In the case of Tambora, the early plinian events in April 1815 had
sustained eruption columns of 33 to 43 km height above the volcano, but the convective columns
during the main ignimbrite phase were probably in the 15 to 20 km range. With estimated magma
source rate of 5.4xl08 kg/s during the 10 April eruption, the source rate of volatiles to the
atmosphere during the is period is calculated as 1 .7xl05 kg/s for sulphur, 4xl05 kg/s for chlorine,
2.7xl05 kg/s for fluorine and about 107 kg/s for magmatic water vapour.

Sulphur emitted by Tambora was initially in the gaseous state, probably dominantly as S02 and
lesser amounts of H2S and OCS, which are the precursor gases to sulphate aerosols and consume
OH radicals. The large mass of magmatic and atmospheric water vapour injected into the
stratosphere during the eruption (2.8xl012 kg) is a major potential source of the OH. Upon
mixing with air, the sulphur dioxide would undergo oxidation to S03 and react with water vapour
in the atmosphere to form an aerosol of sulphuric acid droplets. Reactions of the following type
may account for the conversion of sulphur gases to sulphuric acid aerosol particles in the
atmosphere:

S02 + OH - HOS02 + 02 - S03 + H02

502 + 1/2 02-» S03

503 + H20 - H2S04 (liq)
H2S + 3/2 02 -» H20 + S02
H2S + 202 -> H2S04 (liq)

The above mass estimates of volatile output from the eruption refer to elemental concentration
of sulphur, chlorine and fluorine. Direct analysis of modern volcanic aerosols shows that they
are typically composed of a 75% H2S04 aqueous solution (Hofmann and Rosen 1983). Converting
the above petrologic estimate of 4.3xl010 kg elemental sulphur to sulphuric acid aerosol, we
therefore estimate the Tambora sulphur-rich aerosol mass as 1. 75x10" kg, or an order of
magnitude larger than the 1982 El Chich(5n aerosol (McCormick and Swissler 1983). By
comparison, Hammer et al. (1980) estimate a Tambora volcanic aerosol of 1.5x10" kg on the
basis of the 1816 acidity layer in Greenland ice cores, and Stothers (1984) estimates 2x10" kg
based on observed atmospheric effects. The difference in these estimates is within the
uncertainties of the methods, but several factors make the petrologic estimate a minimum value.
Firstly, further studies of the thickness and distribution of the distal tephra fall deposit preserved
on the ocean floor may conceivably double the total erupted mass estimate and thus double the
estimate of sulphur yield to the atmosphere. Secondly, the Tambora gas emission also involved
about 1x10" kg HC7 and 7.4xl010 kg HF, and the possible involvement of these gases in aerosol
formation cannot be ruled out. Thirdly, the petrologic estimate is only of volatiles exsolved from
the magma at the time of eruption, and does not include a possible separate volatile phase.
Finally, the Tambora stratospheric aerosol or "dust cloud" also contained some particles of
volcanic glass, as demonstrated by the recent identification in a South Pole ice core of Tambora
glass fragments by microprobe analysis (J. Palais, personal communication).

The Halogens

The large-scale introduction of odd-chlorine species into the stratosphere during the 1815
Tambora eruption is important because of the potential of chlorine in catalyzing the removal of
03 and thus damaging the Earth's ozone layer. That layer shields the biosphere from the effects
of damaging solar ultraviolet radiation, such as effects on DNA and the immune-system response,
skin cancer and sunburn. It is generally believed that diffusion of anthropogenic
chlorofluoromethanes (CFC) from the troposphere is currently the principal source of
stratospheric chlorine, but the importance of volcanic emissions as a potential source of
stratospheric chlorine was first pointed out by Stolarski and Cicerone (1974).

Stolarski and Butler (1978) estimated a stratospheric injection rate for volcanic chlorine of
1 .3xl07 kg/yr, or more than three orders of magnitude less than the 10" kg HCf emission during
the 1815 eruption alone. By comparison, the annual release of chlorofluorocarbons is about 7xl08
kg/yr, and the budget of stratospheric chlorine is about 109 kg/yr. HC£ is generally the principal
chlorine molecule in volcanic gases, but studies of the 1980 Mount St. Helens stratospheric cloud
show that concentrations of methyl chloride (CH3Cf) were as high or higher than concentrations
of HCf (Inn et al. 1981). HC£ is highly soluble in water, so possibly large quantities of the
emitted WCt are dissolved in eruption-cloud water and returned to the surface of the Earth as
precipitation during or shortly after eruption.

Although large quantities of chlorine and fluorine are shown to be emitted by Tambora, it should
not be assumed that these gases form aerosols in the stratosphere, as physical and chemical data
indicate that HCf and HF gases are unlikely to form liquid aerosols under normal stratospheric
conditions (Miller 1983; Solomon and Garcia 1984). As shown by Oskarsson (1980) (Figure 8),
however, halogen aerosols may conceivably form at higher temperatures in the eruption column,
and the presence of elevated concentrations of HC7 and HF in volcanic-acidity layers in
Greenland ice cores suggests that halogens have indeed become incorporated into some volcanic
aerosols. Thus, the acidity layer from the 934 A.D. Eldgja eruption in a Greenland ice core
contains at least 65% HC£ (Hammer 1980). Herron (1982) has also shown high levels of both
Ct and F in another Greenland ice-core layer from this eruption. Similarly, Herron (1982) and
Hammer (1977) have both noted elevated CI levels in the Greenland ice-core acidity layer from
the 1783 Laki eruption. Finally, very high CI concentration in a northwestern Greenland ice
core, which was attributed by Herron (1982) to early nineteenth century volcanic activity, may
conceivably represent material from the Tambora eruption. The. ice-core data thus suggest that
Cf and possibly F may enter the volcanic aerosol. This may not imply the formation of a discrete
halogen aerosol, but rather that HC7 and HF may be absorbed and dissolved in the sulphuric acid
aerosol.

HC^ is inert toward ozone, but reaction of HCf with OH leads to formation of atomic chlorine,
followed by the catalytic decomposition of the ozone by the CL Thus in the stratosphere, Ct can
be released from HCf by reactions of the type:

HC£ + OH -* H20 + ce

Similarly, methyl chloride can produce atomic chlorine by photolytic decomposition and attack
by OH. Several reactions involving gaseous chlorine have the effect of converting odd-oxygen
molecules (including ozone) to diatomic oxygen by C?0 catalysis. They are reactions of the type:

03 + C£ -> 02 + CtO

o -t- ao - d2 + a

The only attempt to model the effects of large volcanic chlorine emission on the ozone layer was
made by Stolarski and Butler (1978), who concluded that a Krakatau-size emission, involving
3xl08 kg CtK would result in about 7% depletion of the ozone layer. Chlorine output was two
orders of magnitude higher than this value during the Tambora eruption, and major ozone
depletion cannot be ruled out. Given the great importance of the ozone layer to the biosphere and
climate, the modelling of the potential impact on atmospheric chemistry by a Tambora-size
eruption is timely.

Nothing is known about the possible atmospheric or environmental effects of the large (7.4xl010
kg) HF gas emission during the eruption indicated by our petrologic study. In general, HF is
assumed to be very inert in the stratosphere. The photolysis of HF is shielded by oxygen, and
the reaction of HF with OH is endothermic, so that it is believed that F atoms do not play the
same role in stratospheric chemistry as chlorine atoms (Sze 1978). Furthermore, fluorine and to
some extent chlorine, are known to adsorb onto tephra particles and thus may be rather rapidly
removed from the atmosphere in the tephra fallout (Rose 1977; Oskarsson 1980).

AMBIENT TEMPERATURE

Figure 8: Evolution of volcanic volatiles within an explosive eruption column, showing volcanic volatile
reaction zones (from Oskarsson 1980). In the salt formation zone A, aerosol salt particles are
formed at magmatic temperatures during high-temperature degassing of magma in the vent
region. At temperatures in the range 338° to 700°C, surface adsorption of halogen gases
occurs as they react with silicate material (adsorption zone B). At temperatures below 338°C
sulphuric acid condenses as an aerosol in the condensation zone C.

Fate of Volatiles in the Eruption Column and Atmosphere

An explosive volcanic eruption represents a rapid transfer of heat and mass into the Earth's
atmosphere, resulting in a major thermal and chemical perturbation. In the case of the Tambora
eruption, the thermal energy release alone was equivalent to about 1.3xl027 ergs, most of which
was introduced into the atmosphere over a period of about three or four days. Most of this energy
was expended in convective mixing of the eruption column with ambient air and heating of the
entrained air, resulting in the buoyant rise of the eruption column to heights of 43 km, as a
mixture of pyroclastic fragments, volcanic gases and humid tropospheric air.

Observations and theory (Sparks and Wilson 1982) shows that the solid particle weight fraction
in high-eruption columns (l-nc) is only of the order 0.018; the remainder being almost entirely
entrained atmospheric air and expanding volcanic gases. Assuming that most of the tephra that
generated the fallout deposit (5.8xl013 kg) had entered the lower stratosphere, the mass of
associated air lofted to the stratosphere would then be about equal, and equivalent to
approximately 7xl013 m3 at the surface. The water content of saturated air at 1 atm and 14°C is
about 0.01 kg H20/kg air. Thus the total mass of atmospherically-derived water entrained into
the stratospheric eruption column could have been as high as 5x10" kg. A portion would
condense with rise in the eruption column and cause precipitation, but some would enter the
stratosphere. Although large, this figure is only one-third of the mass of magmatic water
introduced into the atmosphere (1.7xl012 kg), as discussed previously. Normally the content of
water vapour decreases with height due to lowering of both temperature and saturation vapour
pressure and condensation. However, water vapour is likely to be introduced to high levels under
the conditions of elevated temperatures and turbulence within a buoyantly rising eruption column.

Water vapour introduced to the stratosphere by an eruption column could be a major source of
OH radicals by reaction of water vapour with photodissociated oxygen atoms. Evidence from
ground-based spectroscopic measurements of OH during the 1982 El Chich6n eruption indicates
that water vapour was injected at the level of 20 ppm (two to four times normal), and may have
been responsible for the large ozone depletion observed in 1982-1983 (Burnett and Burnett 1984).
Elevated levels of volcanically-derived OH from Tambora may have played a major role in
generation of H2S04 by reaction with S02, in the regeneration of free Cf atoms from HC7 and
in direct reactions with stratospheric ozone.

The field evidence indicates that during the main ignimbrite phase of the Tambora eruption,
transport to the atmosphere was effected by two processes: the eruption column rising above the
centre of the volcano, and secondary eruption columns rising from the coastline around the
volcano as hot pyroclastic flows entered the ocean and flashed seawater to steam. About 35% of
the erupted products entered the ocean in this manner and contributed to the secondary columns,
probably resulting in a coastal ring of composite eruption columns around the entire Sanggar
Peninsula, some 50 km in diameter.

Evidently, extremely variable conditions exist in the eruption column, with great range in
temperature, and mixing proportions of ambient air, condensed water vapour, volcanic gases and
pyroclasts. Temperatures will range from magmatic (about 950°C) to stratospheric air (-60°C).
The fate of the volcanic gases in the eruption column depends on temperature-dependent reactions
in the atmosphere, and although conditions can clearly be highly variable, Oskarsson (1980) has
recognized three zones (Figure 8).

Salt Formation Zone

A spontaneous non-equilibrium degassing occurs during a rapid pressure drop such as an
explosive eruption. In the hottest core of the eruption column, within the jet-like mixture of
pyroclasts and volcanic gases, aerosol salt particles are formed at near-magmatic temperatures.
These are solids condensing from a magmatic gas phase and the solid reaction products of
magmatic gas and its surroundings (Oskarsson 1980). Major sources for the salts are alkali
metals, calcium, aluminum and silica from the silicate melt, and the reactive gases S02, HCf ,
HF and NH3. The dominant products in the salt formation zone are chlorides, fluorides and
sulphates of calcium and the alkali metals. Owing to the high vertical mass flow rates, particles
formed in this zone are likely to represent a small fraction of total aerosol production, and they

will be transported as suspended load high into the eruption column and downwind from the

volcano.

Surface Adsorption Zone

As shown experimentally (Oskarsson 1980), the halogen gases react with silicate ash by surface
adsorption and condensation of the gas phase at temperatures below 700°C. The reactions of the
halogen gases with the glassy tephra will produce components such as calcium fluorosilicates,
sodium and calcium chlorides and sodium fluoride. Halogens adsorbed on tephra particles in this
zone will be removed relatively quickly from the eruption column during fallout. Thus, Rose
(1977) has demonstrated that 17% of the CI released in the 1974 Fuego eruption was stripped
from the eruption plume by adsorption onto tephra particles.

Experiments and observations of the 1970 Hekla eruption show that a large fraction of the
fluorine is stripped from the high-temperature region of the eruption column (338 to 700°C) by
adsorption onto tephra and thus incorporated in the fallout deposit near source (Oskarsson 1980).
Most of the Tambora fluorine emission may have been removed by this process, leading to
fluorosis and thus accounting for the observed death of livestock. During the 1970 Hekla eruption
in Iceland, fluorine-rich fallout led to poisoning of large numbers of livestock up to 200 km from
the volcano (Thorarinsson and Sigvaldason 1972). The tephra fall from the eruption was
unusually rich in adsorbed fluorine (up to 2,000 ppm). The concentration of the adsorbed fluorine
in the fallout deposit was directly dependent on surface area of the tephra grains, and thus the
concentration increased with decreasing grain size. The total mass of fluorine deposited is
estimated as 3xl07 kg, corresponding to 700 ppm of the total erupted mass from Hekla
(Oskarsson 1980).

Condensation Zone

As temperature in the eruption column falls below 338°C, sulphuric acid can condense as an
aerosol by a process controlled by the rate of oxidation of S02 by atmospheric oxygen and
reaction with water vapour. Below 120°C the halogen acids condense and may form an aerosol
prior to condensation of water. The sulphuric acid aerosol droplets can act as a medium in which
other acid components, such as HF, HC7 and water vapour can be dissolved. In the presence of
tephra particles, a portion of the condensed aerosol can be stripped with fallout from the eruption
plume by adsorption onto the silicate ash. Rose (1977) estimates that up to 33% of the sulphur
released by the Fuego 1974 eruption was removed from the atmosphere in this manner.

Effect of the Sulphuric Acid Aerosol on Climate

Pollack et al. (1976) have shown that the optical properties of tephra are distinct from those of
volcanic aerosols such as sulphuric acid, derived from conversion of volcanic gases. The
importance of this was demonstrated during the 1980 Mount St. Helens eruption, when it was
observed that the atmospheric cloud was composed dominantly of a sulphuric acid aerosol a few
days after the eruption. During this eruption, the causes for the relatively short atmospheric
residence time of even the finest-grained tephra were discovered to be due to silicate particle
aggregation (Carey and Sigurdsson 1982). Because of these effects, apparently the potential
climatic impact of a volcanic eruption is not primarily governed by the degree of explosivity or
the volume of erupted magma, but more importantly by the chemical composition of the magma.
Thus, recent studies indicate that the climatological effects of volcanic aerosol emission from
large basaltic fissure eruptions may in fact be more important than the effects of explosive
eruptions of silicic magmas (Sigurdsson 1982).

It is generally accepted that the remarkable global meteorological and optical phenomena,
observed months and years after the Tambora eruption, had a strong connection with activity of
the volcano (Figure 9). Most of these phenomena can be attributed to the effect of the
stratospheric volcanic aerosol. Owing to the sparse meteorological data available, the annual
deviation of the global mean temperature due to the eruption is not well known, but spotty data
indicate a minimum deviation in 1816 of -0.7°C in the northern hemisphere (Stothers 1984). In
a reconstruction of long time series of temperature data from the eastern United States (Landsberg
et al. 1968), the great climatic anomaly of the year 1816 is a unique event that also persists in
1817 (Figure 10). Summer temperature was about 1.5°C below the 200-year average, and the
June 1816 temperature about 3°C below average.

1.8

i i i i i i

▲

• NAKED-EYE STARS

1.4 -

/ Sept \
"/ 1815 N

♦ DIM STARS
■ DARK LUNAR ECLIPSE
A RELATIVE ICE ACIDITY
w ABSOLUTE ICE ACIDITY

1.0 .

\ Tnn#» 1 81 £

^ j une ioio

0.6

\ 19 Sept
X. 1817

0.2

10 April
1 1815

June
1816

■

■ i i i i

0 12 3 4

TIME FROM THE ERUPTION (YEARS)

Figure 9: Change in excess visual extinction (in astronomical magnitude units) following the 1815
Tambora eruption at northern latitudes (after Stothers 1984).

Figure 10: Observed climatic response following the Tambora 1815 eruption. Upper curve is annual
summer temperature data for the eastern United States, at the latitude of Philadelphia,
Pennsylvania, based on several long temperature series. The solid horizontal line shows the
224-year average summer temperature (after Landsberg et al. 1968). Lower curve is annual
June temperature data for New Haven, Connecticut. The lower horizontal solid line shows the
145-year New Haven June mean temperature (World Weather Records 1927).

Devine et al. (1984) and Palais and Sigurdsson (1989) evaluated the possible effect of volcanic
eruptions on climate, and proposed a relationship between the mass yield of sulphur to the
atmosphere from an eruption and the observed decrease in mean northern hemisphere surface
temperature in the one to three years following the eruption, on basis of published temperature
data (Figure 11). Palais and Sigurdsson (1989) found that the mean surface temperature decrease
was related to the estimate of sulphur yield by a power function (r=0.92), with the power to
which the sulphur mass is raised being equal to 0.308. Although these results appear to confirm
a relationship between volcanic sulphur aerosol formation and climatic change, we emphasize that
the temperature deviations are associated with large errors.

As pointed out by Eddy (1988; this volume), the Tambora eruption was coincident with a
depression in solar activity between about 1790-1830, i.e., the Dalton Minimum or the Little
Maunder Minimum in sunspot numbers and aurorae (Figure 13). During these decades the
characteristic 1 1-year cycle in solar activity persists, but the amplitude is reduced by an order of
magnitude or more (Siscoe 1980). Variations in sunspot frequency have been linked to changes
in the solar "constant", and in turn related to climatic changes. Thus the great reduction in
surface temperature on Earth between 1650 and 1730 ("the Little Ice Age") corresponds to the
Maunder Minimum, when there was a sudden reduction in sunspot numbers, almost to zero. It
therefore appears likely that climate was already deteriorating by the beginning of the nineteenth
century, due to reduction in solar activity. This climatic trend was then greatly amplified by the
impact of the Tambora volcanic aerosol, culminating in the "year without summer" in 1816. Both

AO18 data on ice cores and northern hemisphere decadal temperature trends support the contention
that a climatic change had set in by the first decade of the nineteenth century (Figure 12). Thus,
for example, evidence from Peruvian ice cores shows that the decade 1810-20 is characterized
by the most negative AO18 values (coldest temperatures) of the entire record (Figure 12),
culminating in the southern hemisphere wet season of 1819-20 (Thompson etal. 1986; Thompson
and Mosley-Thompson, this volume). The relative contribution of solar variability versus volcanic
aerosol to the deterioration occurring after 1815 is unknown, but John A. Eddy (personal
communication) estimates that solar variability may account for at most 10 to 50%.

SULFUR YIELD (GRAMS)

Figure 11: The observed relationship between sulphur yield to the atmosphere during large volcanic
eruptions and the northern hemisphere temperature decline following the event. Sulphur data
are from Devine et al. (1984) and Palais and Sigurdsson (1989). Climatological data are from
Rampino and Self (1982) and other sources cited in the text. The equation describes the best
fit to the data, with a correlation coefficient of 0.92.

o
o
v.
o

o
o

0.5 -i Northern Hemisphere decadal temperature departures
from the 1881-1975 mean

■■■■■■
V: :::VV

16 i Quelccaya core 1 decadal oxygen isotope average
(1880-1980 mean) 1815

1600 1700 1800

Year A D.

1900

Figure 12: Variations in AO18 in Peruvian ice cores and northern hemisphere surface temperature trends,
showing surface temperature decline in progress before the onset of the Tambora 1815 eruption
(after Thompson et al. 1986).

SUNSPOT NUMBERS NUMBER OF DAYS WITH AURORAE

1780 1800 1820 1840 1860 1880

Figure 13: Auroral and sunspot trends from 1780 to 1880, showing the "Dalton Minimum" or "Little
Maunder Minimum" in solar activity between 1790 and 1830 (after Siscoe 1980). (A) Sunspot
numbers and number of days per year on which aurorae were recorded in Norway. (B) Sunspot
and auroral data in the United States and Europe, south of 54°N Latitude.

Conclusions

The dynamics of the Tambora 1815 eruption columns and source rates of magma and volatiles
can be determined by studying the deposits and petrology of the products. The initial plinian
eruption of Tambora on 5 April was a brief but highly energetic event with eruption rate of
l.lxlO8 kg/s producing a column height of 33 km. In the early phase of the paroxysmal eruption
on 10 April, a plinian column rose to 43 km, with eruption rate of 2.8x10s kg/s. The buoyant
column was only sustained for about three hours, before column collapse occurred due to
increasing eruption rate. Subsequent ignimbrite-phase activity during a three-day period was at
rates of about 5.4xl08 kg/s, producing a co-ignimbrite fall deposit of 5.8xl013 kg and a
pyroclastic flow deposit of 8.2xl013 kg, or a total deposit of 1.4xl014 kg. The convective column
above the volcano during the main ignimbrite phase was at least 20 km high, judging from grain-
size data of the deposit, and thus injected material into the lower stratosphere. Although the ash
fallout affected a broad area, the dispersal was dominantly to the west of Tambora, over Java and
as far as Sumatra. This spread of the eruption plume is consistent with 10-year average rawin-
sonde data for Surabaja in eastern Java, which shows dominant easterly upper troposphere and
lower stratosphere winds for the spring months, with mean velocities ranging from 5 to 10 m/s.
The column height evidence indicates that only about 2% of the erupted mass was emplaced into
the middle stratosphere, up to 43 km, and that the vast majority of the erupted products were
injected in the lower stratosphere and upper troposphere.

The Tambora magma was enriched in volatile components, with 2 to 2.4 wt.% H20, 570 ppm
sulphur, 2,220 ppm chlorine and 570 ppm fluorine. Judging from the difference in volatile
concentration in glass inclusions and in matrix glasses of the tephra, the yield of sulphur to the
atmosphere was 4.3xl010 kg, 10n kg of chlorine, and 7xl010 kg fluorine. Magmatic water
evolved from the volcano was about 2.8xl012 kg, whereas the mass of atmospheric water
entrained in the eruption columns is estimated at 5x10" kg. Source rates of the volatile species
were about 1.7xl05 kg/s for sulphur, 4xl05 kg/s for chlorine, 2.7xl05 kg/s for fluorine and 107
kg/s for magmatic water.

Generation of the sulphuric acid aerosol by gas to particle conversion was probably greatly
facilitated by OH radicals in the eruption cloud, derived dominantly from reactions between
excited atomic oxygen and magmatic water vapour. Assuming a typical volcanic aerosol
composed of 75% H2S04 and 25% water, the petrologic data indicate a minimum Tambora
aerosol mass of 1.75x10" kg. This compares closely to aerosol estimates based on the ice-core
acidity layer (Hammer 1980) and atmospheric phenomena (Stothers 1984).

The very high proportion of halogens released by the Tambora eruption is typical of volcanic
activity of such trachytic magmas in subduction-zone environments. The fate of volcanic halogens
in the atmosphere is unclear at this stage. Fluorine and chlorine most likely form HCf and HF
gas molecules upon degassing from the magma. The latter is relatively inert in the stratosphere,
as HF photolysis is shielded by oxygen and HF is also relatively indifferent to OH abundance.

Chlorine was probably also removed in significant amounts from the high-temperature region of
the Tambora eruption column by adsorption onto tephra. Studies of the 1974 Fuego eruption
indicate that up to 17% of the chlorine was removed by this process (Rose 1977). Although HCt
is not known to form stratospheric aerosols, chlorine may conceivably enter other aerosol
droplets. Studies of ice cores cited above indicate that acidity layers from some eruptions contain
significant chlorine, requiring incorporation of this species into the aerosol by some process.
While HCf is relatively inert in trie stratosphere, reaction with OH or by photolytic reactions

leads to formation of atomic chlorine. As the Tambora eruption cloud was dominantly in the
region below 30 km, which is photolytically inactive, formation of C? and C£0 by the latter
process would have been minor. On the other hand, we contend that water vapour was injected
in large quantities, involving both magmatic and atmospheric water. Thus, OH radicals were
abundant in the eruption column and available for reaction with HC7 to produce atomic chlorine.
Reactions of atomic chlorine with ozone are catalytic, and a single chlorine atom may destroy
thousands of ozone molecules before it becomes inert and enters the HCf reservoir. Independent
of their role in generation of single chlorine atoms, OH radicals from the eruption cloud would
also lead directly to ozone destruction.

The dominant environmental effects of the Tambora eruption were therefore probably of three
types: (1) formation of a sulphuric acid aerosol, leading to a northern hemisphere temperature
reduction of at least 0.7 °C at the surface and stratospheric heating; (2) adsorption of fluorine onto
tephra, leading to very high fluorine levels in the fallout on the ground in Indonesia and resulting
in widespread fluorosis; and (3) extensive ozone depletion as a consequence of generation of odd
chlorine atoms and high levels of volcanically-derived stratospheric OH radicals.

Acknowledgements

This research was carried out with funding from the National Science Foundation (grants EAR-
8607336 and EAR-88041 17), and field studies in Indonesia were made possible by funding from
the National Geographic Society (grant NGS 3390-86). We thank the Volcanological Survey of
Indonesia for collaboration in the field, and the Indonesian Research Council (LIPI), for
permission to undertake research in Sumbawa. The assistance of David Browning in electron
microprobe and grain-size analysis is gratefully acknowledged.

References

Anderson, A.T. 1974. Chlorine, sulfur, and water in magmas and oceans. Geological Society of
America Bulletin 85:1483-1492.

Bach, W. 1976. Global air pollution and climatic change. Review of Geophysics and Space
Physics 14:429-474.

Berresheim, H. and W. Jaeschke. 1983. The contribution of volcanoes to the global atmospheric
sulfur budget. Journal of Geophysical Research 88:3732-3740.

Burnett, C.R. and E.B. Burnett. 1984. Observational results on the vertical column abundance
of atmospheric hydroxyl: description of its seasonal behaviour 1977-1982 and of the 1982
El Chich(5n perturbation. Journal of Geophysical Research 89:9603-9611.

Carey, S.N. and H. Sigurdsson. 1982. Influence of particle aggregation on deposition of distal
tephra from the May 18, 1980 eruption of Mount St. Helens volcano. Journal of
Geophysical Research 87:7061-7072.

_. 1987. The eruption of Vesuvius in A.D. 79, temporal variations in column height and

magma discharge rate. Geological Society of America Bulletin 99:303-314.

Carey, S.N. and R.S.J. Sparks. 1986. Quantitative models of fallout of pyroclasts from explosive
eruptions. Bulletin of Volcanology 48:109-125.

Devine, J., H. Sigurdsson, A. Davis and S. Self. 1984. Estimates of sulfur and chlorine yield
to the atmosphere from volcanic eruptions and potential climatic effects. Journal of
Geophysical Research 89:6309-6325.

Eddy, J. A. 1988. Before Tambora: The Sun and Climate in 1790 to 1830. In: The Year Without
A Summer? Climate in 1816. C.R. Harington (ed.). An International Meeting Sponsored
by the National Museum of Natural Sciences, Ottawa, 25-28 June 1988. Program and
Abstracts, p. 21.

Gorshkov, G.S. 1959. Gigantic eruption of the volcano Bezymianmy. Bulletin Volcanologique
20:77-109.

Hammer, C.U. 1977. Past volcanism revealed by Greenland ice sheet impurities. Nature
270:482-486.

. 1980. Acidity of polar ice cores in relation to absolute dating, past volcanism, and

radioechoes. Journal of Glaciology 25:359-372.

Hammer, C.U., H.B. Clausen and W. Dansgaard. 1980. Greenland Ice Sheet evidence of post-
glacial volcanism and its climatic impact. Nature 288:230-235.

Herron, M.M. 1982a. Impurity sources of F, C£, N03 and S04 in Greenland and Antarctic
precipitation. Journal of Geophysical Research 87:3052-3060.

. 1982b. Glaciochemical dating techniques. In: Nuclear and Chemical Dating Techniques.

American Chemical Society Series 176:303-318.

Hofmann, D.J. and J.M. Rosen. 1983. Stratospheric sulfuric acid fraction and mass estimate for
the 1982 volcanic eruption of El Chichtfn. Geophysical Research Letters 10:313-316.

Inn, E.C.Y. J.F. Vedder, E.P. Condon and D. O'Hara. 1981. Gaseous constituents in the plume
from eruptions of Mount St. Helens. Science 211:821-823.

Johnston, D.A. 1980. Volcanic contribution of chlorine to the stratosphere: more significant to
ozone than previously estimated? Science 209:491-493.

Junghuhn, F. 1850. Java. Volume 2. pp. 1249-1264.

Katsui, Y. 1959. On the Shikotsu pumice fall deposit; special reference to the activity just before
the depression of the Shikotsu caldera. Bulletin of the Volcanological Society of Japan
2(4):33-48.

Landsberg, H.E., C.S. Yu and L. Huang. 1968. Preliminary reconstruction of a long time series
of climatic data for the eastern United States. Report II: On a Project Studying Climatic
Changes. University of Maryland, Institute for Fluid Dynamics and Applied Mathematics,
Technical Note BN-571. 31 pp.

McCormick, M.P. and T.J. Swissler. 1983. Stratospheric aerosol mass and latitudinal distribution
of the El Chichtfn eruption cloud for October 1982. Geophysical Research Letters 10:877-
880.

Miller, E. 1983. Vapor-liquid equilibria of water-hydrogen chloride solutions below 0°C. Journal
of Chemical and Engineering Data 28:363-367.

Neeb, G.A. 1943. The Snellius Expedition in the Eastern Part of the Netherlands East Indies
1928-1929. Volume 5. Brill, Leiden, pp. 55-268.

Oskarsson, N. 1980. The interaction between volcanic gases and tephra: fluorine adhering to
tephra of the 1970 Hekla eruption. Journal of Volcanology and Geothermal Research
8:251-266.

Palais, J. and H. Sigurdsson. 1989. Petrologic evidence of volatile emissions from major historic
and pre-historic volcanic eruptions. In: Understanding Climate Change. A. Berger,
R.E. Dickinson and J.W. Kidson (eds.). American Geophysical Union, Washington, D.C.,
Geophysical Monograph 52:31-53.

Pannekoek van Rheden, J.J. 1918. Geologische Notizen uber die Halbinsel Sanggar, Insel
Sumbawa. Zeitschrift fur Vulkanologie 4:185-192.

Petroeschevsky, W.A. 1949. A contribution to the knowledge of the Gunung Tambora
(Sumbawa). KoninklijkNederlands AardrijkskundigGenoostschap, Tijdschrift 66:688-703.

Pollack, J.B., O.B. Toon, C. Sagan, A. Summers, B. Baldwin and W. Van Camp. 1976.
Volcanic explosions and climatic change: a theoretical assessment. Journal of Geophysical
Research 81:1071-1083.

Raffles, S.H. 1835. Memoir of the Life and Public Services of Sir Thomas Stamford Raffles.
Volume I. Duncan, London. 267 pp.

Reclus, E. 1871. The Earth. G.P. Putnam and Sons, New York. 666 pp.

Rose, W.I. 1977. Scavenging of volcanic aerosol by ash: atmospheric and volcanologic
implications. Geology 5:621-624.

Sapper, K. 1917. Katalog der Geschichtlichen Vulkanausbruche. Truber, Strasbourg. 358 pp.

Self, S., M. Rampino, M. Newton and J. Wolff. 1984. Volcanological study of the great
Tambora eruption of 1815. Geology 12:659-663.

Self, S. and J. A. Wolff. 1987. Comments on "The petrology of Tambora volcano, Indonesia: a
model for the 1815 eruption" by J. Foden. Journal of Volcanology and Geothermal
Research 31:163-170.

Sigurdsson, H. 1982. Volcanic pollution and climate: the 1783 Laki eruption. EOS 63:601-602.

Sigurdsson, H., J.D. Devine and A.N. Davis. 1985. The petrologic estimation of volcanic
degassing. JOkull 35:1-8.

Sigurdsson, H. and S. Carey, (in press). The 1815 eruption of Tambora, Indonesia: generation
of co-ignimbrite ash fall during entrance of pyroclastic flows into the ocean. Bulletin of
Volcanology.

Siscoe, G.L. 1980. Evidence in the auroral record for secular solar variability. Review of
Geophysics and Space Physics 18:647-658.

Solomon, S. and R.R. Garcia. 1984. On the distribution of long-lived tracers and chlorine species
in the middle atmosphere. Journal of Geophysical Research 89:1 1,633-1 1,644.

Sparks, R.S.J. 1986. The dimensions and dynamics of volcanic eruption columns. Bulletin of
Volcanology 48:3-15.

Sparks, R.S.J, and L. Wilson. 1982. Explosive volcanic eruptions - V. Observations of plume
dynamics during the 1979 Soufriere eruption, St. Vincent. Geophysical Journal of the
Royal Astronomical Society 69:551-570.

Stoiber, R.E., S.N. Williams and B. Huebert. 1987. Annual contribution of sulfur dioxide to the
atmosphere by volcanoes. Journal of Volcanology and Geothermal Research 33:1-8.

Stolarski, R.S. and D.M. Butler. 1978. Possible effects of volcanic eruptions on stratospheric
minor constituent chemistry. Pure and Applied Geophysics 117:486-497.

Stolarski, R.S. and R.J. Cicerone. 1974. Stratospheric chlorine: a possible sink for ozone.
Canadian Journal of Chemistry 52:1610-1615.

Stolper, E.M. and J.R. Holloway. (in press). Experimental determination of the solubility of
carbon dioxide in molten basalt at low pressure. Earth and Planetary Science Letters.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science 224:1191-
1198.

Sze, N.D. 1978. Stratospheric fluorine: a comparison between theory and measurements.
Geophysical Research Letters 5:781-783.

Thompson, L.G., E.M. Thompson, W. Dansgaard and P.M. Grootes. 1986. The Little Ice Age
as recorded in the stratigraphy of the tropical Quelccaya Ice Cap. Science 234:361-364.

Thorarinsson, S. and G.E. Sigvaldason. 1972. The Hekla eruption of 1970. Bulletin
Volcanologique 36(2):269-288.

Verbeek, R.D.M. 1885. Krakatau. Batavia, Indonesia. 495 pp.

Van der Broeck. 1834. Oosterling 1:183.

Walker, G.P.L. 1980. The Taupo pumice: products of the most powerful known (ultraplinian)
eruption? Journal of Volcanology and Geothermal Research 8:69-94.

Zollinger, H. 1855. Besteigung des Vulkanes Tambora auf der Insel Sumbawa und Schilderung
der Erupzion Desselben in Jahr 1815. J. Wurster (ed.). Winterthur. 20 pp.

The Possible Effects of the Tambora Eruption in 1815 on Atmospheric
Thermal and Chemical Structure and Surface Climate

R.K.R. Vupputuri1
Abstract

A coupled 1-D radiative-convective-photochemical diffusion model that takes into account the
influence of ocean inertia on global radiative perturbation is used to investigate the possible
climatic and other atmospheric effects of the Tambora eruption of 1815. The volcanic cloud was
introduced in the model stratosphere between 20-25 km, and the global average peak aerosol
optical thickness was assumed to be 0.25. Both the aerosol optical thickness and aerosol
composition determining the optical properties were allowed to vary in the model atmosphere
during the life cycle of the volcanic cloud. The results indicate that the global average surface-
temperature decreases steadily from the date of eruption in 1815 with maximum cooling of 1°K
occurring in spring 1816. The calculations also show significant warming of the stratosphere,
with temperature increasing up to 15°K at 25 km in less than six months after the date of
eruption. The important effects of the Tambora eruption on stratospheric ozone and UV-B
radiation at the surface are also discussed.

Introduction

Modelling studies of global radiative perturbations caused by volcanic eruptions are extremely
important to understand the nature of past and present changes in the Earth's climate. As pointed
out by Kondratyev (1983), the primary mechanism by which volcanic activity influences the net
radiation (and consequently the climatic system) is through alterations in the aerosol content of
the atmosphere. It is well known that volcanic aerosols scatter and absorb solar radiation and
absorb and emit infrared radiation. The net effect causes general cooling of the troposphere and
the surface, and warming in the stratosphere. The extent of the cooling and warming, however,
depends upon the composition, size distribution and the morphological structure of aerosols.

Several prominent volcanic eruptions took place during the past 200 years (Laki, 1783; Krakatau,
1883; Mount Agung, 1963; and El Chichtfn, 1982). The largest and deadliest was that of Mount
Tambora in April 1815, on the island of Sumbawa, Indonesia (8°S, 118°E). It was also the
world's greatest ash eruption since the end of the last ice age. The dust veil index (a measure of
increase in the atmospheric turbidity arising from small particles injected into the stratosphere)
has been estimated to be more than twice that of Agung (Lamb 1970; Robock 1981, Mitchell
1982). The Tambora eruption is blamed by some studies in the literature for the cold summer of
1816 on the east coast of North America, where average temperature was the lowest on record
with 1.5 to 2.5°C below the seasonal norm (Landsberg and Albert 1974). Indeed 1816 was called
"the year without a summer" in New England and eastern Canada, where daily minimum
temperatures were abnormally low from late spring through early autumn. It was also very cold
and wet in western Europe in the summer of 1816, although it was milder at some stations in
eastern Europe. Despite these earlier claims of strong cooling on a regional basis, the more recent
analysis of climatic data for 1781-1983 by Angell and Korshover (1985) suggests that there is no

Canadian Climate Centre, 4905 Dufferin Street, Downsview, Ontario M3H 5T4, Canada.

clear evidence of strong cooling on a hemispheric basis following the Tambora eruption. As
pointed out by Angell (1988), the reason for the lack of evidence of strong volcanically-induced
cooling on a hemispheric basis is that such cooling may or may not have been observed
depending upon the extent of sea-surface warming in the eastern equatorial Pacific due to an El
Nino event following the volcanic eruption.

It is clear from the above arguments that the injection of ash, sulphur and dust into the
stratosphere by a large volcano could alter the existing radiative-photochemical balance of the
Earth's atmosphere, in turn leading to changes in the vertical temperature and chemical structure
and surface climate. In this respect a volcanic event as large as the Tambora eruption provides
a unique opportunity for a case study of the response of the climatic system to a global radiative-
photochemical perturbation. It also allows testing of our ability to model and understand the
nature of the climatic system and climatic change. Several model calculations have been made
in the past to study the climatic impact of Agung and El Chichdn eruptions (Hansen et al. 1977;
Robock 1984; McCracken and Luther 1984; Vupputuri and Blanchet 1984). All these calculations
showed warming of the stratosphere and cooling in the troposphere and at the surface, although
the amplitudes of warming and cooling differ depending upon the assumed peak aerosol optical
depth, altitude of peak aerosol concentration and optical properties. In this paper a 1-D time
dependent radiative-convective-photochemical diffusion model (RCPD model) taking into account
the thermal inertia of oceans is used to investigate the thermal and chemical response of the
atmosphere and surface climate to radiative-photochemical perturbations caused by the Tambora
eruption in 1815.

The Climatic Model

The coupled one-dimensional climatic model extending from the surface to 60.5 km is described
in detail in Vupputuri (1985). It involves combining the radiative-convective model of the type
developed by Manabe and Wetherald (1967) with a photochemical transport model. Starting from
the assumed vertical temperature distribution and chemical composition, the basic procedure is
to compute the local net radiative heating and cooling and photochemical sources and sinks at
each altitude to determine the vertical temperature and trace-constituent structure with the time
marching method. The upward heat transfer by atmospheric motions is taken into account
implicitly by a simple numerical procedure called convective adjustment - first introduced by
Manabe and Strickler (1964). Using this numerical procedure, the vertical lapse rate is restored
to a pre-assigned stable value (6.5° km"1) whenever it becomes greater due to radiative heat
transfer. The relative humidity is fixed in the model, and the mixing ratio of water vapour is
computed as a function temperature. Cloud-top altitude is assumed to be at 6.5 km, with a
fractional cloud cover kept at 50%. The climatic model is coupled to the underlying surface
through the energy balance equation at the surface. For climatic change calculations in this study,
thermal heat capacity of the underlying surface is assumed to zero for the land and the value
appropriate for the upper mixed layer in the case of the ocean. For several other computational
details of the model, see Vupputuri (1985).

The Radiative Transfer Model

The solar radiation code used to compute the short wave solar heating within the atmosphere is
based on the delta-Eddington method, which is computationally efficient and fairly accurate
(Joseph et al. 1976). It considers the absorption and scattering by atmospheric gases (H20, C02,
03 and N02), aerosols and cloud droplets. To compute the infrared cooling due to H20, 03 and

C02, the analytical formulae for the mean transmissivities of finite frequency intervals derived
by Kuo (1977) have been adopted. The mean transmission functions take into account the
temperature effect and overlapping absorption between gases, and the computed transmissivities
have been shown to be in good agreement with line-by-line calculations. For other trace gases,
such as N20, CH4 and CFCs, empirical expressions for the mean band absorptivities based on
laboratory and spectroscopic data (Burch et al. 1962; Ramanathan et al. 1985) were adopted.
Also considered in long-wave calculations are the heating or cooling-rate contributions due to
aerosols in the atmospheric window band. Both IR and solar radiation codes have been validated
by comparing them with other standard radiation codes through participation in the workshop on
the intercomparison of radiative codes in the climatic models (World Meteorological Organization
1984).

The Photochemical Model

The photochemical system considered here includes the important reactions affecting the
concentration of ozone and other relevant trace constituents in the atmosphere above 10 km: they
are listed in Table 1. The chemical species considered are: Ox (O, O('D), 03), HOx (H, HO,
H02), NOx(N, NO, N02, HN03), CH4, N20, CF2C£2, CFC£3 and Ctx(Cl, C(0, CfN03)
chemical species. The concentrations of H20, CH4 and H2 are specified based on observations.
The chemical kinetics and photochemical data used are based on NASA (1985) recommendations.
Computed photodissociation rates take into account the effects of Rayleigh scattering and
absorption and scattering by aerosols and cloud droplets. The chemistry of the global troposphere
below 10 km is considered to be much more complex, and therefore to have more uncertainties
than in the stratosphere due to the presence of higher hydrocarbons, heterogeneous processes and
long photochemical relaxation times for the chemical species. In view of some of these
uncertainties, the chemistry is frozen in the troposphere by prescribing the ozone concentration
below 10 km for the purpose of this study.

The Perturbed Aerosol Model for the Tambora Eruption

To calculate possible climatic and other atmospheric effects of the Tambora eruption,
observational information on perturbed aerosol concentration and optical properties as a function
of time starting from the date of eruption are needed. It is not, however, possible to obtain such
detailed information for Mount Tambora. For this study it is assumed that the dust veil index for
Mount Tambora is roughly twice that of Agung. Using the peak optical depth of 0.125 as a
representative global average value for the added aerosols in the case of Agung (Hansen et al.
1978), the maximum optical thickness for Mount Tambora is estimated to be 0.25. The assumed
shape of the vertical aerosol profile producing this maximum optical thickness is shown in
Figure 1. The vertical distribution of perturbed aerosols and the altitude of peak aerosol
concentration are similar to those of Agung and El Chichtfn eruptions. Since the added aerosol
concentration and the optical properties are expected to vary with time during the lifetime of the
volcanic cloud, it is not realistic to assume fixed optical thickness and properties for the model
calculations. In the present calculations, both the aerosol optical thickness and properties are
varied as a function time starting from the date of volcanic eruption. The assumed variation of
perturbed aerosol optical thickness is illustrated in Figure 2. For the first four months after the
eruption, the perturbation optical thickness is allowed to increase linearly to a maximum 0.25,
and during this period ash optical properties are assigned for the added aerosols. From four to

Table 1: The Principle Chemical and Photochemical Reactions Used in the Model.

O i U ii O _1_ o

KJ2 + ny u 1 u

OH 4_ O -» HO 4- O
Un + KJ-j ^ nUi + KJj

KJ -f- vj^ 1 ivl * 1 IVI

OI4 4- O _» R 4- O
KJli 1 KJ f 1 1 W2

O _L_ O O _1_ O

U -r U3 -» U2 + U2

Ufl 4_ O _» OR 4- 00

rHJ2 + U3 ^ iJrl t Z.KJ2

u3 -t- np ^ u2 + u

R 4- O 1 Xyt > RO 4. \A

ri + \J2 ivi * nUi + ivi

u3 + np -* u2 + u(^ dj

14 4. 0 OU 4. 0
rl + KJj —* Url + L>2

D( U) + U3 -* U2 + U2

I40 _l_ UO 14 O 4. 0

MU2 + riU2 ~* rt2U2 + U2

U( JJj + M^U + M

R O 4. In. — > OR 4. OR

INUj + np -* INU + U

14 O 4- OI4 _» T4 O 4- 14 O

IN U t U t IVI ^ IN U2 ' IVl

OR 4- 14 O _» R O 4- O

Url t nUj ^ iItvJ + kj2

MO 1 O _» MO a. O
INW2 INU + U2

RO 4- MO 4- \A -» RMO 4- \A
rlKJ + IN KJ2 t IVI ^ 1IINVJ3 T IVI

MO 1 O _» MO a_ O
INU -r U3 -* INU2 + U2

OR 4- RMO »1 1 O 4- MO

Url + rl in Uj ^r^u t INU3

MO _i_ K 1. _» M 1 O
1N2U + nc -* IN2 + u

RO 4- MO — » OR 4- MO
tlUj -r 1NU ^ Url + INU2

M O -L O/1 r»\ _» M 1 O
IN2U -r \Jy U) -* IN2 + U2

OP CP 1 U ,, » /~<p CO < CP

M O J- o^1 _» MO J- MO
IN 2U > '-A L> ,) ^ IN U T IN U

fcpi) 1 U„ CVCP 4- PP

MO -i_ l-i.. _» M _i_ O
INU + nf -* IN + KJ

cp 1 t t ur^ 1 it
+ H2 ~* rll^r + ri

M -L O _» MO u_ O
IN t INU T U

CP 4- PR _ * VXCP 4- PR
+ Url4 -* rlCt -r C1I3

M _u MO _» M _i_ O
IN -r IN KJ ^ IN 2 < KJ

CP 4- 14 0 up p 1 0

L-t + rlU2 ^ rlL.t + U2

u n 1 o^' n^i _» oh _l 014

MtU t U^ 1J J — » Url -r Uo

OT4 4. RPC _» R O 4- CP

un + net ^ 1I2U + k^i

R O -I- h.. — » H 4- OI4
ri 2 w t np -* n -r un

UC 0 1 O ^ OR J- CP
tiK^i t U - * Url t

fH 4_ (V'Til — » OH 4- PH

RPP 4- l-i 1, _» R 4. OP

fU _l O -1- M -» PR O 4- M

\— n3 1 KJ2 1 ivi v^- n j 2 1 ivi

PC I O n PPn 4- O
V_ X. -|- U3 ^ LtU -r U2

PH O 4- MO — » PH O 4- MO
V-1I3W2 1 IN U ^ ^rljVj 1 IN KJ7

CPr\ 1 O t CP 4- O
HO "T U ^ ^ t T U2

ru O 4- O _» T4 PO 4- HO

PPr> 4- MO — » OP 4. MO
LIO -r IN KJ ^ K^l t IN U2

H:CO + \u> -» H + HCO

Cfo + N02 + M -» Cf N03 + M

HCO + 02 - CO + H02

C£N03 + hp - CfO + N02

Un + LU -• CU2 + H

CtN03 + HCf -* C( 2 + HNO3

H2 + 0('D) - H + OH

CfN03 + O -> Cf 0 + N03

HNO3 + hf -* OH + N02

10 months the peak optical thickness remains the same but the optical properties are changed
from ash to sulphuric acid. After 10 months the optical thickness decreases exponentially until
it reaches the background stratospheric value while the optical properties change from sulphuric
acid to background stratospheric aerosols. The optical parameters (extinction coefficients, single
scatter albedo, asymmetry factors) for ash, sulphuric acid and background stratospheric aerosols
vary as a function of wavelength both in solar and infrared spectra. Sulphuric acid properties
chosen for this study are those reported in Bundeen and Fraser (1982), and they are derived
assuming the aerosol particles are composed of 75% H2S04 and 25% H20. The ash optical
properties are determined by assuming an imaginary refractive index of 0.002 (Patterson and
Pollard 1983). Both the sulphuric acid and ash properties are similar to those used for the El
Chichdn volcanic eruption.

Results and Discussion

Before discussing the results of atmospheric response to radiative-photochemical perturbations due
to the Tambora eruption, it should be pointed out that for the prescribed annual average insolation
and background stratospheric aerosols, the coupled 1-D model produced reference atmosphere

simulations of minor trace constituents and temperature that are representative of natural
background atmosphere in tropical latitudes. A detailed comparison of reference atmosphere
model simulations of ozone and temperature, and discussion on some of the deficiencies in 1-D
model calculations, were given in Vupputuri (1985).

SO.O-i

CONCENTRATION PROFILE OF RER0S0LS
45'°" GIVING MAX. OPTICAL THICKNESS

40.0-

Figure 1: Assumed aerosol concentration profile (NO/CC) for Mount Tambora volcanic eruption which
produces the maximum optical depth of 0.25.

Effects on Solar and Infrared Radiation

Figure 3 shows the calculated change in direct, diffuse and net solar radiation at the surface as
a function of time beginning from the date of eruption of Tambora in 1815, while the
corresponding effects on infrared flux and planetary albedo are illustrated in Figure 4. Direct
solar flux decreases by about 15% following the eruption (Figure 3). However this decrease is
compensated in large measure by an increase in diffuse radiation, leaving a net decrease of solar
flux at the surface of about 4%. There are no observational data on direct solar radiation
following the Tambora eruption. The visual extinction curve constructed by Stothers (1984)
suggests that the excess zenithal visual extinction increases rapidly during the first four to five
months from the eruption date, and then returns to normal gradually within four to five years.
The time variation of visual extinction is quite consistent with the variation of calculated direct
solar radiation following the eruption of Tambora. As indicated in Figure 4, the infrared flux at
the surface also decreases by up to 4%, while planetary albedo increases by about 7% following
the Tambora eruption.

0.15

ASSUMED VARIATION OF AEROSOL OPTICAL DEPTH

FOR TAMBORA ERUPTION

1617 1818

TIME AFTER ERUPTION! YEARS)

ure 2: The assumed variation of perturbed stratospheric aerosol optical thickness with time following
the Mount Tambora eruption in 1815.

C 15.0

5 C9

S en

-5.0-

EFFECT OF TAMBORA ERUPTION ON SOLAR RADIATION

DIRECT FLUX
DIFFUSE FLUX
TOTAL FLUX

1616

1817 1818

TIME AFTER ERUPTIONIYEARS)

;ure 3: The calculated change in direct, diffuse and net solar radion (in %) at the surface following the
Mount Tambora eruption.

Figure 4: The calculated percentage change in the infrared flux at the surface and planetary albedo
following the Mount Tambora eruption.

Effects on Global Climate

Figure 5 shows the calculated response of stratospheric and surface temperature as a function of
time following the Tambora eruption. The response is shown for two different assumed heat
capacities for the underlying lower-boundary surface. The solid curve represents the calculated
surface-temperature response assuming that the lower boundary has no heat capacity (valid
assumption for the land surface). The dashed curve, on the other hand, corresponds to the heat
capacity of the underlying surface equivalent to that of 70 m of ocean water. The combined
response of land and ocean is given by the dotted curve. In the stratosphere, the ocean heat
capacity has no effect on temperature response due to volcanic forcing caused by the Tambora
eruption (dash-dot curve in Figure 3). Note however that the stratosphere responds much more
quickly to the global radiative perturbation due to volcanic forcing. As indicated in Figure 5, the
eruption of Tambora could have resulted in a peak warming of about 15°K within less than six
months after the eruption. But in the case of the troposphere and the surface it takes almost twice
as long to reach the maximum cooling. The calculated maximum cooling for the land surface
following the Tambora eruption is about 2°K, and for the combined land and ocean it is roughly
1°K. It may be seen from Figure 5 that not only the land surface cools faster than the ocean
surface but it also warms faster than ocean as the temperature recovers to its pre-volcanic state.

15.0

10.0

5.0

THE EFFECT OF TAMBORA ERUPTION ON GLOBAL CLIMATE

STRATOSPHERIC WARMING (25 KM)

LJ
CD

en
ce

LJ
Q_
II
LJ

-1.0-

-2.0

TIME AFTER ERUPTION! YEARS)

' ■ ■_' ' ' ' 1 '

1818 1820 1821 J822 JB23---r-.ifl£tr.r. "7.3825

SURFACE COOLING

NO SURFACE HEAT CAPACITY

SURFACE HEAT CAPACITY 1 70M OF OCEAN)

COMBINED LAND AND OCEAN HEAT CAPACITY

-3.0J

Figure 5: The calculated stratospheric and surface-temperature change (O K) following the Tambora
eruption. The surface-temperature response is shown for two different assumed heat capacities
for the underlying lower boundary surface.

The physical explanation for the calculated temperature effects in the present model is
straightforward. For the aerosol concentrations and optical properties assumed in this study, the
net effect of the added aerosols in the stratosphere from the Tambora eruption would be to
increase the planetary albedo and decrease both the solar and thermal radiation at the surface
(Figures 3, 4). This leads to cooling in the troposphere and at the surface. The local heating in
the stratosphere, on the other hand, is caused by both the absorption of thermal radiation
emanating from the warmer lower atmosphere and in situ absorption of solar radiation (in the
near infrared and UV part of the spectrum) by the added aerosols. Due to low air density in the
stratosphere, only a small change in radiational energy is needed to cause a large change in local
air temperature.

As pointed out earlier, although surface temperatures were abnormally low in the summer of
1816 (following the Tambora eruption) in New England, eastern Canada and parts of western
Europe, the evidence for a strong land-surface temperature cooling on a global average basis (as
indicated by the results of this study) following the Tambora eruption is rather weak. As Angell
and Korshover (1985) indicate, the extent of surface cooling on an hemispheric or global basis
depends critically upon the timing and strength of El Nino in relation to the time of a large
volcanic eruption. Indeed Quinn et al. (1978) have found some evidence for a moderate El Nino
one year after the Tambora eruption. This might partially explain the reason for the lack of
evidence of strong cooling on a global or hemispheric basis in 1816 following the Tambora
eruption.

Effects on Stratospheric Ozone and UV-B Radiation at the Surface

As mentioned earlier, the absorption of solar and thermal radiation by the added aerosols in the
stratosphere from a large volcanic eruption such as Tambora can lead to a large increase in the
stratospheric temperature. The altered temperature in turn effects the concentration of ozone and
other minor constituents through temperature-dependent reaction-rate coefficients. Due to the
inverse relationship between ozone and temperature in the middle stratosphere, the temperature
increase in that region causes ozone concentration to decrease. Ozone is also destroyed in the
stratosphere by enhanced photodissociation resulting from the backscattered UV radiation from
the added aerosols. Figure 6 shows the computed ozone column reduction and UV-B radiation
increase at the surface following the Tambora eruption. It is seen from Figure 6 that the added
aerosols in the stratosphere from the Tambora eruption could have resulted in up to about 7%
decrease in total ozone, which translates into up to 15% increase in the UV-B radiation at the
surface following the eruption event. Although there were no observational data following the
Tambora eruption to verify the computed ozone depletion, the analysis of Umkehr observations
following El Chich6n in 1982 by DeLuisi et al. (1984) clearly indicates the evidence of
volcanically-induced ozone depletion. This lends support for the theoretical calculations shown
in Figure 6.

EFFECT OF

TAMBORA ERUPTION ON ATMOSPHERIC OZONE
AND SURFACE UV-B RADIATION

TOTAL OZONE DEPLETION

CHANGE

• 8

SURFACE UV-B RADIATION

RCENTAGE

.0-

LJ
Q_

-4

-8

.0-

1815

1816

> • i i •

1817

TIME HETER

' ■ ' i i

1818 1819 1820

ERUPT I ON I YEARS)

Figure 6: The calculated changes in the total ozone column and UV-B radiation at the surface (in %)
following the Mount Tambora eruption.

Concluding Remarks

The Tambora eruption in 1815 - the largest and deadliest eruption in recorded history - also
injected the greatest amount of ash, sulphur and dust into the stratosphere. The event that
produced the largest dust veil index provided a unique opportunity to investigate climatic and
other atmospheric response to global radiative perturbation, and to understand the effects of
volcanic eruptions on past and present climate.

Despite the simplicity of a 1-D radiative-convective-photochemical diffusion model that does not
include the interaction of radiative heating perturbation with atmospheric dynamics and other
uncertainties regarding the input data, the magnitude of land-surface temperature decrease
generally agrees with the observed cooling in the east coast of North America, where 1816 was
called "the year without a summer". However there is lack of strong evidence from the
observational data to support the computed combined land-ocean surface cooling on hemispheric
or global bases. The computed cooling may or may not have been observed depending upon the
timing and extent of the El Nino event following the Tambora eruption. There were no
observational data in 1816 to verify the computed changes in ozone and temperature in the
stratosphere following this eruption. However, the well-documented observational evidence of
temperature warming (Quiroz 1983) and ozone depletion (DeLuisi et al. 1984) in the stratosphere
following the El Chichtfn eruption lends support for the calculated changes in the case of the
Tambora eruption.

One should exercise caution in interpreting either the observations or the model results presented
here. Temperature and ozone observations are not as detailed as desired for accurate
determination of observed climatic and total ozone changes on a global average basis. Further,
the observed climatic change is not a part of other natural events such El Nino, QBO and the sun-
spot cycle, or simply noise in the climatic system (Hansen et al. 1978). Climatic calculations in
the model are too simplified and, in particular, the model is not capable of handling the complex
interactions between radiative heating and large-scale dynamics and other cloud feedback
mechanisms. Nevertheless, the calculated amplitudes of climatic and ozone perturbations resulting
from Tambora's eruption are large enough to believe that volcanic eruptions do indeed strongly
affect the Earth's climate and the ozone layer. As pointed out by Angell (1988), the reason that
the evidence for volcanically-induced cooling of the Earth's surface in the past was so uncertain
is that the cooling may or may not have been observed depending upon the extent of warming
due to El Nino following the volcanic eruption. Detailed observations and careful data analysis
taking into account the impact of other natural events following a major eruption such as
Tambora would enable us to understand better the role of volcanic aerosols in altering the
radiative-photochemical balance of the global atmosphere and climate.

Acknowledgements

The author thanks: Dr. G.J. Boer and the Director General of the Canadian Climate Centre for
encouragement and support; Mr. Frank Szeckeli and Lynda Smith for programming and
manuscript preparation support, respectively.

References

Angell, J.K. 1988. Impact of El Nino on the delineation of tropospheric cooling due to volcanic
eruptions. Journal of Geophysical Research 93:3696-3704.

Angell, J.K. and J. Korshover. 1985. Surface-temperature changes following the six major
volcanic episodes between 1780 and 1980. Journal of Climate and Applied Meteorology
24:937-951.

Bundeen, W.R. and R.S. Fraser. (eds.) 1983. Radiative effects of the El Chichtfn volcanic
eruption: preliminary results concerning remote sensing. NASA Technical Memorandum
84959, Goddard Space Flight Center, Greenbelt, Maryland.

Burch D.E., D. Grynak, E.B. Singleton, W.L. France and D. Williams. 1962. Infrared
absorption by C02, H20 and minor atmospheric constituents. AFCRL-62-68%, Ohio State
University Contract AF19(604)-2366.

DeLuisi, J.J., C.L. Mateer and W.D. Komhyr. 1985. Effects of the El Chichdn stratospheric
aerosol cloud on Umkehr measurements at Mauna Loa, Hawaii. Atmospheric ozone.
C.S. Zerefos and A. Ghazi. (eds.). D. Reidel Publishing Co., Dordrecht.

Hansen, J.E., W.C. Wang and A. A. Lacis. 1978. Agung eruption provides test of a global
climate perturbation. Science 199:1065-1068.

Joseph, J.H., W.J. Wiscombe and J. A. Weinman. 1976. The delta-Eddington approximation for
radiative flux transfer. Journal of Atmospheric Sciences 33:2452-2459.

Kondratyev, K. Ya. 1983. Volcanoes and Climate. R.D. Bojkov and B.W. Boville (eds.). World
Meteorological Organization, WCP-54.

Kuo, H.L. 1977. Analytic infrared transmissivities of the atmosphere. Beitrage zur Physik der
Atmosphaere 50:331-349.

Lamb, H.H. 1970. Volcanic dust in the atmosphere, with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society of London
A226:425-533.

Landsberg, H.E. and J.M. Albert. 1974. The summer of 1816 and volcanism. Weatherwise
27:63-66.

Manabe, S. and R.F. Strickler. 1964. Thermal equilibrium of the atmosphere with a convective
adjustment. Journal of Atmospheric Sciences 21:361-385.

Manabe, S. and R.T. Wetherald. 1967. Thermal equilibrium of the atmosphere with a given
distribution of relative humidity. Journal of Atmospheric Sciences 24:241-259.

McCracken, M.C. and F.M. Luther. 1984. Preliminary estimates of the radiative and climatic
effects of the El Chich(5n eruption. Geofisica Internacional 23(3):385-401.

Mitchell, J.M., Jr. 1982. El Chichdn, weather-maker of the century. Weatherwise 35:252-259.

NASA Panel for Data Evaluation. 1985. Chemical kinetics and photochemical data for use in
stratospheric modelling. California Institute of Technology, Jet Propulsion Laboratory,
Publication 85-37, Pasadena, California. 217 pp.

Patterson, E.M. and CO. Pollard. 1983. Optical properties of the ash from El Chich<5n volcano.
Geophysical Research Letters 10:317-320.

Quinn, W.H., D.O. Zopf, K.S. Short and R.T.W. Kuo Yank. 1978. Historical trends and
statistics of Southern Oscillation, El Nino and Indonesian droughts. Fishery Bulletin
76:663-678.

Quiroz, R.S. 1983. The isolation of stratospheric temperature change due to the El Chichdn
volcanic eruption from non-volcanic signals. Journal of Geophysical Research 88:6773-
6780.

Ramanathan, V., H.B. Singh, R.J. Cicerone and J.T. Kiehl. 1985. Trace gas trends and their
potential role in climate change. Journal of Geophysical Research 90:5547-5566.

Robock, A. 1981. A latitudinally dependent volcanic dust veil index and its effect on climate
simulations. Journal of Volcanology and Geothermal Research 11:67-80.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science 224:1191-
1198.

Vupputuri, R.K.R. 1985. The effect of ozone photochemistry on atmospheric and surface
temperature changes due to increased C02, N20, CH4 and volcanic aerosols in the
atmosphere. Atmosphere-Ocean 23:359-374.

Vupputuri, R.K.R. and J. -P. Blanchet. 1984. The possible effects of El Chichtfn eruption on
atmospheric thermal and chemical structure and surface climate. Geofisica Internacional
23(3):433-447.

World Meteorological Organization. 1984. The intercomparison of radiation codes in climate
models (ICRCCM). Longwave clear-sky calculations. F.M. Luther (ed.). World
Meteorological Organization, WCP-93.

Climatic Effects of the 1783 Laki Eruption

Charles A. Wood1

Abstract

From 8 June 1783 to 7 February 1784, 12 km3 of lava poured from a series of volcanic vents in
southern Iceland, devastating farmland and ultimately causing a severe famine that decimated the
island's human and animal populations. This Laki eruption apparently had much more widespread
consequences, however, for its ash and sulphurous gases were transported in the lower
atmosphere across Europe causing a remarkably warm summer, which was followed in Europe,
eastern North America and at least some parts of Asia by one of the most severe winters on
record. Poor weather continued through the summer and winter of 1784. Scarce meteorological
measurements and abundant written records and proxy data graphically document these climatic
anomalies.

Conventional volcano-climate theories cannot readily explain these apparent climatic effects. Great
eruptions such as Tambora, 1815 or Krakatau, 1883 explosively emplace volcanic aerosols into
the stratosphere where, during a two- to three-year period before they are finally flushed out,
they absorb incoming radiation, thus depriving the lower atmosphere of a portion of its heat. For
the Laki eruption there is no direct evidence that significant quantities of sulphuric aerosols
reached the stratosphere. If the continuing climatic deterioration of 1784 was related to the clearly
volcanic weather of 1783, then a new mechanism needs to be identified.

Introduction

The first recognition that volcanism may effect climate was Benjamin Franklin's (1784)
speculation that the non-explosive Laki eruption in Iceland could be responsible for the poor
climate in Europe and North America during the summer and winter of 1783. Although the Laki
eruption was the largest effusive volcanic activity in historic times, it, and its possible climatic
effects, have been studied only at the reconnaissance level. Nonetheless, general information on
the nature and chronology of the eruption, and compilation of reports of anomalous weather
thereafter, provide strong support for Franklin's prescient theory, and lead to puzzling aspects
of conventional eruption-climate relationships.

Laki Eruption of 1783

The only accessible accounts of the Laki eruption are Thorarinsson's (1969) 20-year old
preliminary report and a recent abstract by Thordarsson et al. (1987), which incorporate
eyewitness descriptions and reconnaissance geological mapping. Most of the 1783 activity
occurred along a 25-km-long series of fissures, creating large fields of lava flows with small
cones marking their vents. The eruption began on 8 June 1783, following three weeks of
earthquakes, and continued for eight months (until 7 February 1784). Thorarinsson estimates that
~ 12 km3 of tholeiitic lava flows were emplaced, with the majority (10 km3) produced during the
first 50 days (8 June to 28 July 1783). From examination of Thorarinsson's map (Figure 1)

NASA Johnson Space Center, SN2, Houston, Texas 77058, U.S.A.

showing the extent of the lava at different dates, it appears that a significant portion of the flows
had formed by 21 June and the majority of the 10 km3 by 22 July.

Figure 1: Map of Laki from Thorarinsson (1969). The Laki lavas of 1783 are lightly stippled.

Thus, the extrusion rate during the first two weeks may have been considerably greater than the
50-day average of 2,200 nrVsec estimated by Thorarinsson. Following this initial, high-rate
extrusion, activity continued at a much slower rate for the next six months. The total area
covered by the flows is 565 km2 (Thoroddsen 1925), which implies an average thickness of —20
m.

A recent development has been the recognition that the Laki fissure activity was part of volcanic-
tectonic eruption centred on the Grimsvotn caldera, northeast of Laki (Thordarsson et al. 1987;
Thordarsson and Self 1988). During and after the Laki fissure eruptions, Grimsvotn had a series
of explosive eruptions between 18 July 1783 and 26 May 1785.

From the perspective of climatic effects it is important to know how much explosive activity
occurred at the beginning of the eruption. According to eyewitnesses the eruption was very
violent during the first few days, with enormous lava fountains. Thorarinsson pointed out that
groundwater may have been involved in some of these eruptions because phreatomagmatic tephra
cones were formed. Based upon his field measurements of buried ash from Laki, Thorarinsson
believed the volume of explosively erupted material was 0.3 km3; he discounted earlier estimates
of 3 km3. Thordarsson et al. give a similar value (0.21 km3) for the tephra.

Effects of the Laki Eruption in Iceland

Thorarinsson (1969) states that the Laki eruption was the greatest catastrophe in Icelandic history,
and the mortality statistics bear him out. Gases from the eruption stunted the growth of grass so
that it was insufficient to feed livestock. As a result, 50% of the cattle, 79% of the sheep, and
76% of the horses starved. During the next three years 24% (9,000 people) of the human
population died of starvation, and the population did not return to earlier levels until 1780
(Jackson 1982).

Ogilvie (1986) cites contemporary diaries that provide graphic information on the effects of the
eruption. From 8 June to at least 26 August, "the air was full of ash and smoke. On the rare
occasions that we have had a glimpse of the sun it has looked like the reddest blood". Grass
turned yellow and white due to "sulphurous rain" and it withered to the roots. Fishermen were
not able to go to sea because of "murmurings" (earthquakes?) and continuous smoke that reduced
visibility to less than a mile. In northeastern Iceland, one writer recorded, "From early June, and
to this time (13 August) we have lived in continual smoke and fog, sometimes accompanied by
sulphur-steam and ashfalls."

Weather in Iceland during the eruption seemed variable from location to location according to
written records analyzed by Ogilvie (1986), but she concludes that 1783-84 winter began very
early, and was very severe and long-lasting. Various accounts state that the ground was frozen
with hard ice from 2 October until the end of April 1784. All fiords were reported to be frozen
over in late February 1784 (for the first time in 39 years), and sea ice was very widespread and
long-lasting. The summer of 1784 was also cold and wet, with occasional periods of frost and
sleet. Ogilvie's summary of seasonal weather shows that there was uniformly cold weather across
Iceland for five seasons (summer 1783 through summer 1784) after the onset of the Laki
eruption. Also, 1782 was unusually cold, although apparently not as uniformly so as following
the eruption.

Haze and Dust

The tremendous quantity of volcanic gases and dust released during the Laki eruption was
reported from many locations in the northern hemisphere. The English rector Gilbert White
(1977) wrote:

The summer of the year 1783 was an amazing and portentous one, and full of
horrible phenomena... the peculiar haze, or smoky fog, that prevailed for many
weeks in this island, and in every part of Europe, and even beyond its limits, was
a most extraordinary appearance, unlike anything known within the memory of
man. ...The sun at noon looked as blank as a clouded moon, and shed a rust-
coloured ferruginous light... and was particularly lurid and blood-coloured at rising
and setting.

Icelandic accounts describe the volcanic haze from early June to the end of August (Ogilvie
1986), and Lamb (1970) reported the following first sightings of dry fog or haze in Europe and
eastward:

Copenhagen 29 May

France 6 June

North Italy 18 June

Syria 1 July

Altai, central Russia 1 July.

Evidently, the dry fog spread eastward and southeastward at an average rate of approximately
250 km/day during the first month of the eruption. This is only about 10% of the rate of
propagation for the Krakatau haze (2,700 km/day; Russell and Archibald 1888). The difference
in velocity may be due to the differences in direction (east for Laki, west for Krakatau) or the
altitude (tropospheric for Laki, stratospheric for Krakatau). Benjamin Franklin (1784) also noted
that the haze was seen in North America, although the original sources and details of this
observation were not reported. Nonetheless, the haze persisted long enough, or rose to different
atmospheric heights with differing wind directions, so that it was carried both eastward to Europe
and westward to North America.

Volcanic dust also fell out of the sky in Europe. Lamb (1971) reported that tulips in Holland
were damaged by the dust and sulphurous smells during 18-24 June 1783. In Scotland the dust
was thick enough to destroy crops in June. The detection of sulphurous odours in Europe proves
that the haze was volcanic and not from some unknown forest fire, for example. The odour and
eye irritation imply that the haze was at low altitudes. Volcanic dust that fell in Holland 1 1 days
after the eruption started, apparently was transported much more slowly than dust from other
Icelandic eruptions. Ash from the 1875 eruption of Askja reached Europe within a day
(Thorarinsson 1963).

The Summer of 1783

White's (1977) account quoted above continues:

All the time the heat was so intense that butchers' meat could hardly be eaten on
the day it was killed; and the flies swarmed so in the lanes and amid hedges that
they rendered the horses half frantic, and riding irksome.

Instrumental temperature records reveal that 1783 was the warmest English July on record
(Kington 1978). Other early thermometer data for six other major European cities allow
quantification of how extreme the summer heat was in 1783. World Weather Records data
(Figure 2) for Stockholm, Copenhagen, Edinburgh, Berlin, Geneva, and Vienna for a 31 -year
period centred on 1783 demonstrate that July 1783 was 1.6 to 3.3°C warmer than the 31-year
average. In general the amount of the July temperature anomaly is closely correlated with the
distance of each city from Laki (Figure 3). Thus, however the haze raised the summer
temperatures in Europe, the effect was most pronounced where the haze was thickest, and the
excess heating declined where the haze was probably less intense. Temperature data from eastern
North America (Landsberg et al. 1968) reveal that the summer of 1783 was significantly hotter
than the 225-year average (Sigurdsson 1982). Figure 4 shows the same data graphically.

^ 22 n 1

^ 16 ~l ■ 1 1 1 1 1 ■ 1 ■

1778 1780 1782 1784 1786 1788

Year

Figure 2: Average July temperatures (data from World Weather Records) for six European cities
(Stockholm, Copenhagen, Edinburgh, Berlin, Geneva, and Vienna) for the seven-year period
centred on 1783, the year of Laki's eruption. The 31-year average is based on recorded
temperatures for the 31 years centred on 1783.

O 4

Laki to City Distance in Km

Figure 3: Deviation of July 1783 temperatures from 31-year averages as a function of the distance from
Laki to six European cities. Edinburgh's anomaly is less than expected based on the other
cities, suggesting that the temperature increases were not latitudinally uniform. Perhaps these
anomalies - from only one month after the start of the eruption - were due to tropospheric dust,
which would not be as uniformly distributed as stratospheric aerosols.

There are various sources of proxy weather information for this period; e.g., in Switzerland the
summer was drier as well as warmer than normal (Pfister 1981), and there was a drought and
poor harvest in Finland (Schove 1954).

Additional circumstantial evidence that the summer of 1783 was warm includes a severe drought
in the Yangtze region of China (Wang and Zhao 1981). The Yangtze drought continued into
1784, but in both years there were floods in southeastern China and the Hwang Ho (Yellow)
River Basin. An extraordinarily severe famine throughout Japan in the summer of 1783, however,
was not caused by drought: Mikami (1988; Mikami and Tsukamura, this volume) has shown that
an excess of rain destroyed many crops, and that, in fact, the summer of 1783 was wettest and
coolest in Japanese history. Based on the high price of wheat in Delhi, India, the rains probably
failed in 1783 with a consequent famine (Pant et al. 1988). These extremes in Asian weather
during the summer of 1783 exhibit regional variations in the response to volcanism that are
similar to previously-documented patterns in North America (Lough and Fritts 1987).

E
E

24.5

24.0"

23.5

Sigurdsson, 1982

— 225 yr Average

Variance ■-

1 1 1 1 1 1 1

1782 1783 1 784 1785 1786

Year

Figure 4: Average summer temperatures in the eastern United States in the 1780s compared to the 225-
year average. Data from Landsberg et ctl. (1968) as reported by Sigurdsson (1982).

Winter of 1783-84

Scant temperature measurements, abundant proxy data and anecdotal accounts demonstrate that
the winter of 1783-84 was one of the most severe on record in Europe and North America. The
average January temperature for six European cities was 3° below the 31 -year average centred
on 1784 (Figure 5). Proxy temperature data (from viticulture/agriculture) indicate Switzerland
had two extremely severe winters in 1783-84 and 1784-85, with the first year being the worst
(Pfister 1981). The longest period of sea ice around Iceland also occurred during the winter of
1783-84 when temperatures were nearly 5° colder than the 225-year average (Sigurdsson 1982).
The next two winters were also significantly colder than normal (Figure 6). Information compiled
by Ludlum (1966) includes the following records for the winter of 1783-84 in the eastern United
States:

Longest in early American history (last snow in late April),

Near record depth of snowcover,

Near record low temperatures,

Greatest seasonal snowfall ever in New Jersey,

Longest period of below zero temperatures ever in New England,

Longest freezing ever of Chesapeake Bay,

Longest and coldest winter in Maine,

One of the greatest southern snowstorms (18-19 December),

Freezing of Charleston Harbour (ice skating occurred),

Freezing of Mississippi River at New Orleans (13-19 February 1784),

Ice floes in Gulf of Mexico 100 km south of New Orleans.

CD
Q.

CD
>

1778 1780 1782 1784 1786 1788

Year

Figure 5: Average January temperatures (data from World Weather Records) for six European cities
(Stockholm, Copenhagen, Edinburgh, Berlin, Geneva, and Vienna) for the seven-year period
centred on 1783, the year of Laki's eruption. The 31-year average is based on recorded
temperatures for the 31 years centred on 1783.

225 yr Average —

t ■ r

1780 1781 1782 1783 1784 1785 1786

Year

Figure 6: Average winter temperatures in the eastern United States in the 1780s compared to the 225-year
average. Data from Landsberg et al. (1968) as reported by Sigurdsson (1982).

Ludlum (1966) provides graphic evidence of the severity of the winter by quoting contemporary
letters and newspapers. Following a series of early and frequent storms, the worst weather of the
winter occurred in mid-February, when minimum recorded temperatures for eight nights at
Hartford, Connecticut were about 12°C or colder. From Virginia, James Madison wrote on
11 February 1784, "We had a severer season and particularly a greater quantity of snow than is
remembered to have distinguished any preceding winter." In another letter of 5 March 1784,
George Washington complained that he, "arrived at this Cottage [Mount Vernon, Virginia] on
Christmas eve, where I have been locked up ever since in frost and snow."

The February cold spell froze the western end of Long Island Sound, and at New York City the
Narrows between Staten Island and Long Island were blocked by ice for 10 days, preventing
ships in Manhattan harbours from reaching the sea. Baltimore harbour was frozen by 2 January
1784 and remained closed until 25 March. Chesapeake Bay was nearly completely frozen, and
the Delaware River at Philadelphia froze on 26 December 1783 and was icebound until 12 March
1784. Ludlum reports that even the southern harbour of Charleston, South Carolina was frozen
in February, "having produced ice strong enough for skating on, which is very uncommon there."
The most amazing phenomenon of the winter was the freezing of the Mississippi River at New
Orleans, which Ludlum (1966, p. 154) reports has happened only once before (1899):

On the 13th of February, 1784, the whole bed of the river, in front of New
Orleans, was filled up with fragments of ice, the size of most of which was from
twelve to thirty feet, with a thickness of two to three. This mass of ice was so
compact, that it formed a field of four hundred yards in width, so that all
communications was interrupted for five days between the two banks of the
Mississippi. On the 19th, these lumps of ice were no longer to be seen. "The
rapidity of the current being then at the rate of two thousand and four hundred
yards an hour," says Villars, "and the drifting of the ice by New Orleans having
taken five days, it follows that it must have occupied in length a space of about one
hundred and twenty miles. These floating masses of ice were met by ships in the
28th degree of latitude [in the Gulf of Mexico].

That the unusually cold winter was not just confined to the eastern United States is clear from
the Hudson's Bay Company records indicating 1783-84 had the fifth worse ice blockage of
Hudson Strait on record (Catchpole 1988).

Summer of 1784

Summer temperature in England averaged 0.5°C, and as much as 1.6°C, below the long-term
norm during 1784. The driest 12 months in English history began in August 1784 (Kington
1978). Temperatures in the eastern United States tended to be below average: < 53.9°F (12. 1 °C)
in Philadelphia and <48.0°F (8.9°C) in New Haven (Bray 1978). Tree-rings indicate a marked
growth minimum in the growing season of Douglas fir in Nevada, Utah, and Wyoming in 1784
and 1785 (Woodhouse 1988). Tree-ring densities from the Mackenzie Delta region of Canada
indicate that 1784 had a very cold summer (Parker 1988). Light coloured rings in black spruce
at the treeline near Quebec indicate low temperatures shortened the growing season in 1784
(Filion et al. 1986). Similarly, tree rings from Alaska show that the cool weather of 1784
extended far to the north (Oswalt 1957).

Winter of 1784-85

In Switzerland, the long duration of snowcover during the 1784-85 winter resulted in widespread
growth of the snow mold Fusarium nivale, which led to harvest failure of the spring grain crops
(Pfister 1981). In Bern the winter was also very severe, with snow on the ground for more than
150 days (Pfister 1978). Winter, spring, and early summer of 1784 in Brittany were disastrous,
with a cold winter, hail at the end of April, spring floods, and a drought until the end of June
(Sutherland 1981). In England, 1785 tied with 1674 as the coldest March on record (Kington
1978). The date of freezing of Lake Suwa, Japan (Figure 7), occurred 22 days earlier than the
long-term average (Arakawa 1954).

Subsequent Seasons

1785 was the worst year of the decade in Brittany, and one of the worst of the century. In some
areas no rain fell between January and August (Sutherland 1981). The summer was also cool in
England (Bray 1978), and the autumn of 1786 was one of the three coldest in English history
(Kington 1978).

In the eastern United States, summer temperatures were lower than normal in New Haven in
1785 but returned to normal in 1786 (Bray 1978).

CD ^
™ 0

J- U)
CO OS
LU *-

CO >

>*<

CO _

a I

CD I-

CO 5-

Q CD
N

0 O

40
30-
20-
10

-10-
■20-

■30-
1 7

Arakawa (1954)

1784-5

Feb. 1

Jan.

0
«

N
0
1 0

0
J*
CO

Dec. 1

1780 1790

Year

1800

Figure 7: 23-year record of the date of freezing of Lake Suwa, Japan centred on 1783 (data from
Arakawa 1954). In 1783 freezing occurred 22 days earlier than normal.

Mechanisms to Explain Observed Climatic Anomalies

Each of the unusual weather records reported above can be dismissed as a freak occurrence
within the normal range of variation, and thus not requiring a special origin. Consideration of
the entire list of anomalies - and these are only the items that were found during a brief
examination of secondary and tertiary historic records - suggests, however, that a period of
unusual weather affected various places in the northern hemisphere from the summer of 1783
through 1785. The principal observations to be accounted for are:

Early summer 1783: Dry fog over Europe, western Asia, and the United States
Summer 1783: Hot in Europe, United States and China; cold in Iceland
Winter 1783-84: Exceptionally cold in Europe, United States and Japan
Winter 1784-85: Very cold in Europe and Japan
Summer 1785: Cool and dry in Europe and United States

Benjamin Franklin's famous 1784 communication to the Literary and Philosophical Association
of Manchester was the first suggestion that volcanic eruptions might effect climate (see
Sigurdsson 1982):

During several of the summer months of the year 1783, when the effect of the
sun's rays to heat the earth in these northern regions should have been greatest,
there existed a constant fog over all of Europe, and a great part of North America.
This fog was of a permanent nature; it was dry, and the rays of the sun seemed to
have little effect toward dissipating it, as they easily do a moist fog, arising from
water. They were indeed rendered so faint in passing through it, that when
collected in the focus of a burning glass, they would scarce kindle brown paper.
Of course, their summer effect in heating the earth was exceedingly diminished.

Hence the surface was early frozen.

Hence the first snows remained on it unmelted, and received continual additions.

Hence the air was more chilled, and the winds more severely cold.

Hence perhaps the winter of 1783-84 was more severe, than any that had happened
for many years.

The cause of this universal fog is not yet ascertained. Whether it was adventitious
to this earth, and merely a smoke, proceeding from the consumption by fire of
some of those great burning balls or globes which we happen to meet with in our
rapid course around the sun, and which are sometimes seen to kindle and be
destroyed in passing our atmosphere, and whose smoke might be attracted and
retained by our earth: or whether it was the vast quantity of smoke, long continuing
to issue during the summer from Hecla in Iceland, and that other volcano which
arose out of the sea near that island, which smoke might be spread by various
winds, over the northern part of the world, is yet uncertain.1

Franklin's first speculation, that the summer fog and winter coldness could be due to smoke from
a meteor is rather bizarre, and has been forgotten (except perhaps as an unremembered
contribution to the idea that a comet or asteroid collision with the Earth 65 million years ago
resulted in extinction of the dinosaurs). His second option, that smoke from "Hecla in Iceland,
and that other volcano which rose out of the sea near that island" caused the observed weather
anomalies is more enduring.

The general concept that volcanic activity can affect the climate has been developed since the
obvious weather anomalies following the eruption of Krakatau in 1883 (Russell and Archibald
1888). Mitchell (1961); Lamb (1971); Self et al. (1981); Devine et al. (1984) and others have
demonstrated statistically that years in which volcanic aerosols are ejected into the stratosphere
by volcanic eruptions are typically followed by one to three years of temperatures 0.2 - 0.5 °C
below average. Thus, although there are doubters, there is a widely-promoted model that
explosive eruptions of sulphur-rich magmas can implant sulphuric-acid aerosols into the
stratosphere where they remain suspended for a few years. The aerosols spread around the globe,

Italics added by editor.

absorbing incoming solar radiation, thus heating the stratosphere and, by a reduction in the
amount of radiation reaching the ground, cooling the Earth's surface (Sigurdsson 1982; Rampino
and Self 1984).

The Laki eruption does not appear to fit this general model because it was not a classical
eruption, such as Krakatau or Tambora, which explosively ejected aerosols into the stratosphere.
Effusive eruptions, like Laki and typical of activity at Hawaiian volcanoes, are thought to have
only minimal explosive activity, with nearly all magma flowing quietly across the Earth's surface
as lava flows. Wood (1984a); Devine et al. (1984); and Stothers et al. (1986) proposed, however,
that aerosols from Laki may have entered the stratosphere even though the eruption was largely
quiescent. The last two groups proposed that heat from fire fountains and lava fields could
generate convective plumes that would rise into the stratosphere. This effect would be enhanced
by normal atmospheric mixing across the tropopause which replaces the entire air mass in the
lower and middle stratosphere with tropospheric air every one to two years (Flohn 1968).
Following my previous suggestion (Wood 1984a), this mixing could result, over the prolonged
period of the Laki eruption and with the possibly high altitude of convectively-transported
materials, in substantial deposition of sulphur aerosols in the stratosphere. Thus, the Laki
eruption may have emplaced sufficient material in the stratosphere to produce the multi-year
climatic effects.

Most of the observed climatic effects in the early to mid- 1780s can be explained by the volcanic
hypothesis (Lamb 1970). The dry fog in Europe, the Near East, and North America, and the
sulphurous smells, burning of eyes, and singeing of tulips in western Europe resulted from the
tropospheric movement of Laki dust and sulphur aerosols mainly to the east but apparently also
to the west. Gilbert White reported that the dry fog lasted one month, which coincided with the
period of maximum volcanic activity (Wood 1984a). The hot summer weather in 1783 in Europe,
United States and China is an unusual occurrence; no other volcanic eruption is associated with
such hot weather. Perhaps the heated gases in the mid-troposphere hindered normal convection
so that heat was trapped near the surface. The cold summer in Iceland, however, was presumably
caused by the blockage of sunlight by persistent dense haze and smoke from Laki. Similar
immediate cooling near dense volcanic plumes occurred at Tambora (Rampino, this volume), as
well as Krakatau and Mount St. Helens (10° and 8°C below normal, respectively; reported in
Simkin and Fiske 1983).

The exceptionally cold winter of 1783-84 in Europe, North America and Japan is proposed to
have resulted from the standard volcanic mechanism of stratospheric warming and hence
tropospheric cooling due to the abundance of volcanic sulphuric-acid aerosols in the stratosphere.
The very cold winter in Europe and Japan during 1784-85 and the following (1785) cool and dry
summer in Europe and United States can only be explained if large numbers of aerosols remained
in the stratosphere through 1785. This would be consistent with other large explosive eruptions,
such as Tambora and Krakatau which were followed by lower than normal temperatures for one
to three years (Rampino and Self 1984).

Uncertainties

These explanations would be impossible if aerosols from Laki did not enter the stratosphere. One
significant piece of evidence suggests they did not. Sulphuric acid droplets from volcanic-eruption
clouds fall to the Earth everywhere under the passing cloud, but the existence of the droplets is
recorded only when they fall on permanent ice fields, as in Greenland and Antarctica, where they

leave an acidic trace in that year's ice layer (Hammer 1977). The largest acid spike in the Camp
Century ice core in Greenland occurs in 1783 (Hammer 1977); but as pointed out by Sigurdsson
(1982) there is no acid anomaly for 1784. The 1783 anomaly could be due to either tropospheric
or stratospheric transport of aerosols from Laki (only about 1200 km to the east). An anomaly
for 1784, which could only occur if significant aerosols were stored in the stratosphere for a
year, would be strong evidence that Laki materials reached the stratosphere; the lack of a 1784
anomaly is most consistent with no stratospheric contribution. If this is true then the present
understanding of volcanic influences on climate require that the cold winter of 1784-85 is
unrelated to the Laki eruption and the cold winter of 1783-84.

A second uncertainty is whether Laki was actually the volcano that produced the anomaly
recorded in the Camp Century ice core and that caused the observed climatic effects. One
confusing piece of evidence is the report (in Lamb 1970) that the dry fog was first observed on
29 May 1783 in Copenhagen and on 6 June in France. Yet the Laki eruption is recorded to have
started only on 8 June 1783. Was another, earlier eruption responsible for the dry fog and other
effects?

Eleven eruptions are recorded to have begun in 1783 (Simkin et al. 1981), and two other little
known ones may have occurred (Table 1). In terms of volume, Laki was the largest eruption and
Asama, in Japan was the second largest. All other eruptions of the year are thought to have been
considerably smaller, but one of them may have been important.

Nyey

As Franklin (1784) noted there was another "...volcano which rose out of the sea near..."
Iceland. This volcano was the temporary island of Nyey or Noyoe (New Island) which formed
over the Mid-Atlantic Ridge some 50 km southwest of the Reykanes Peninsula in southwestern
Iceland. Nyey began erupting by 1 May 1783 and produced a large deposit of pumice that floated
on the sea for about 250 km around the volcano, causing great hardship for sailors (Lyell 1969).
By autumn, when the Danish government sent an expedition to lay claim to the island, Nyey had
been destroyed by wave action. One of the few descriptions (and a drawing, Figure 8) of the
eruption is reproduced in Thorarinsson (1967). The Danish Captain Mindelberg of the brig
Boesand first saw a smoke column on 1 May and wrote in his ship log, "At three o'clock in the
morning we saw smoke rising from the sea and thought it to be land; but on closer consideration
we concluded that this was a special wonder wrought by God and that a natural sea could
burn... When I caught sight of this terrifying smoke I felt convinced that Doomsday had come."
(quoted by Thorarinsson 1967). On 3 May Boesand approached the area of the smoke plume, but
..."when we had come within half a mile of the island we had to turn away for fear that the crew
might faint owing to the enormous sulphur stench."

Two, perhaps similar, eruptions near Iceland during the last 25 years provide comparisons. In
1963 the island of Surtsey formed off the southern coast, and in 1973, a new cone and lava flow
was constructed near Surtsey at Heimaey on the Vestmann Islands. Both eruptions were similar
to the account of Nyey in that explosive eruptions produced scoria cones, but both Surtsey and
Heimaey were armoured by lava flows and have been able to withstand wave erosion. Neither
of the recent island eruptions produced as large a pumice field as reported for Nyey, and only
minor amounts of ash fell in Europe. Based on the modern examples, it seems unlikely that Nyey
could have caused the widespread effects commonly attributed to Laki, but the reported
widespread pumice and lack of detailed information makes it impossible to reject completely the
notion that Nyey contributed to the 1783 climatic phenomena.

Ou*y/$L +^//Aa/ <^£<t+*» t/Z* +*rit€c

Figure 8: Drawing and last page of text from Captain Mindelberg's report on the Nyey eruption
southwest of Iceland in May 1783. Reproduced from Thorarinsson (1967).

Table 1: Volcanic Eruptions, 1783.

Start Date

Volcano

Location

VEI

Comment

05 May

Nyey

Off Iceland

09 May

Asama

Japan

Biggest eruption
in August

12 May

Barren Island

Andaman Islands

08 June

Laki

Iceland

till 8 Feb. 1784

? July

Izalco

El Salvador

18 August

Vesuvius

Italy

>3 years

03 September

Sakurajima

Japan

03 December

Iwaki?

Japan

Kurikoma

Japan

Kanaga

Alaska

>3 years

Unnamed

Greenland Sea

Unnamed

North Atlantic

1 From Simkin et al. (1981). ? = Unknown date within 1783; ?? = uncertainty if eruption
occurred in 1783. VEI 4 = Volcanic eruption index (0-8); VEI 4 = 10s to 109 m3 of ejecta.

Asama

The largest historic eruption of Asama volcano in Japan began on 9 May 1783. Bullard (1976)
and others have suggested that this eruption caused the climatic anomalies of 1783. I have
previously summarized (Wood 1984b) recent Japanese literature on the 1783 Asama eruption
which tends to discount it as the source of the dry fogs and other early summer climatic effects.
The main argument is that although eruptive activity began in early May, nearly half of the total
of 0.5 km3 of ejecta was deposited during two days of intense eruptions on 3 and 4 August 1783,
and most of the remainder formed during the next five days (Imai and Mikada 1982), two months
after the dry fogs were reported. Asama may have contributed to the generally cool winter of
1783, but it did not contribute to the strong atmospheric effects of early summer.

Laki

The observation that dry fog was reported in Europe 10 days before the onset of activity of Laki
is well dated by eyewitness accounts. Thordarsson and Self (1988) discovered by studying old
Icelandic maps that the Grimsvotn basaltic caldera, about 50 km northeast of Laki along the
fissure trend, erupted repeatedly throughout the Laki eruption. As proposed by Sigurdsson and
Sparks (1978), activity along the Laki fissure system was probably intimately tied to activity at
the Grimsvotn caldera. Thordarsson and Self (1988) suggest that there may have been an eruption
at Grimsvotn in May, before the first Laki activity. Thus, the Laki/Grimsvotn system may have
produced all the dry fogs of the summer of 1783.

Summary

There are many loose ends in the story of Laki and its possible climatic effects. In this report a
variety of readily available observations of unusual climatic phenomena occurring during the two
years following the eruption is presented. The simplest assumption is that these anomalies are
related to Laki, just as similar types and durations of climatic phenomena are clearly accepted
as being associated with Tambora's eruption in 1815. It is most likely that eruptions of the
Laki/Grimsvotn system caused the dry fogs and hot summer of 1783 and the cold winter of 1783-
84. In order to cause the cold winter of 1783-84 volcanic aerosols must have reached the
stratosphere. And probably the cold winter of 1784-85 was due to the same stratospheric aerosols,
which however, left no trace in the Greenland ice core for 1784. If Laki produced all of these
effects, present volcano-climate models are inadequate to explain how. If Laki was not
responsible, then a major eruption 200 years ago is completely missing from our records.

In compiling the historical data for the 1780s it became obvious that most reports are from the
eastern United States and western Europe. A much greater effort is required to search the
historical (and proxy) archives of Africa, Asia, South and Central America, and central and
western United States to further define possible climatic effects of Laki and other eruptions.

Acknowledgement

I thank Michael Helfert for sharing information concerning the unusual weather following the
eruption Laki.

References

Arakawa, H. 1954. Fujiwhara on five centuries of freezing dates of Lake Suwa in central
Japan. Archtv fiir Meteorologie, Geophysik und Bioklimatologie 6:152-166.

Bray, J.R. 1978. Volcanic eruptions and climate during the past 500 years. In: Climatic Change
and Variability: A Southern Perspective. A.B. Pittock et al. (eds.). Cambridge University
Press, Cambridge, pp. 256-262.

Bullard, F.M. 1976. Volcanoes of the Earth. University of Texas, Austin. 579 pp.

Catchpole, A.J.W. 1988. Presentation at The Year Without A Summer? Climate in 1816. An
International Meeting Sponsored by the National Museum of Natural Sciences, 25-28 June
1988, Ottawa.

Devine, J.D., H. Sigurdsson, A.N. Davis and S. Self. 1984. Estimates of sulfur and chlorine
yield to the atmosphere from volcanic eruptions and potential climatic effects. Journal of
Geophysical Research 89:6309-6325.

Filion, L., S. Payette, L. Gauthier and Y. Boutin. 1986. Light rings in subarctic conifers as a
dendrochronological tool. Quaternary Research 26:272-279.

Flohn, H. 1968. Climate and Weather. McGraw Hill, New York. 256 pp.

Franklin, B. 1784. Meteorological imaginations and conjectures. (Reprinted in Sigurdsson,
1982).

Hammer, C.U. 1977. Past volcanism revealed by Greenland ice sheet impurities. Nature
270:482-486.

Imai, H. and H. Mikada. 1982. The 1783 activity of Asama Volcano inferred from the
measurements of bulk density of tephra (pumice) and the old documents. Bulletin of the
Volcanological Society of Japan 27(l):27-43.

Jackson, E.L. 1982. The Laki eruption of 1783: impacts on population and settlement in
Iceland. Geography 67:42-50.

Kington, J. A. 1978. Historical daily synoptic weather maps from the 1780s. Journal of
Meteorology 3(27):65-70.

Lamb, H.H. 1970. Volcanic dust in the atmosphere: with a chronology and assessment of its
meteorological significance. Proceedings of the Royal Society of London 266:425-533.

Landsberg, H.E., C.S. Yu and L. Huang. 1968. Preliminary reconstruction of a long time
series of climatic data for the eastern United States. Report II: On a project studying
climatic changes. University of Maryland, Institute of Fluid Dynamics, Technical Note
BN-571.

Lough, J.M. and H.C. Fritts. 1987. An assessment of the possible effects of volcanic
eruptions on North American climate using tree-ring data, 1602 to 1900 A.D. Climatic
Change 10:219-239.

Ludlum, D.M. 1966. Early American Winters, 1604-1820. American Meteorological Society,
Boston. 283 pp.

Lyell, C. 1969. Principles of Geology, Vol. 1. Johnson Reprint Collection, New York. pp. 371-
374. (A reprint of 1830-33 edition).

Mikami, T., 1988. Climatic reconstruction in historical times based on weather records.
Geographical Review of Japan 61(1): 14-22.

Mitchell, J.M. 1961. Recent secular changes of global temperature. Annals of the New York
Academy of Sciences 95:235.

Ogilvie, A.E.J. 1986. The climate of Iceland, 1701-1784. Jokull 36:57-73.

Oswalt, W.H. 1957. Volcanic activity and Alaskan spruce growth in A.D. 1783. Science
126:928-929.

Pant, G.B., B. Parthasarathy and N.A. Sontakke. 1988. Climate over India during the first
quarter of the nineteenth century. In: The Year Without A Summer? Climate in 1816. An
International Meeting Sponsored by the National Museum of Natural Sciences, 25-28 June
1988, Ottawa. Abstracts p. 52

Parker, M.L. 1988. Presentation at The Year Without A Summer? Climate in 1816. An
International Meeting Sponsored by the National Museum of Natural Sciences, 25-28 June
1988, Ottawa.

Pfister, C. 1978. Fluctuations in the duration of snow cover in Switzerland since the late 17th
century. Danish Meteorological Institute, Climatological Papers 4:1-6.

. 1981. An analysis of the Little Ice Age climate in Switzerland and its

consequences for agricultural production. In: Climate and History. T.M.L. Wigley,
M.J. Ingram and G. Farmer (eds.). Cambridge University Press, Cambridge, pp. 214-
248.

Rampino, M.R. and S. Self. 1984. Sulphur-rich volcanic eruptions and stratospheric aerosols.
Nature 310:677-679.

Russell, F.A.R. and E.D. Archibald. 1888. On the unusual optical phenomena of the
atmosphere, 1883-1886, including twilight effects, coronal appearances, sky haze,
coloured suns, moons. (Reprinted in T. Simkin and R.S. Fiske [1983 pp. 397-404]).

Schove, D.J. 1954. Summer temperatures and tree-rings in North Scandinavia, A.D. 1461-
1950. Geografiska Annaler 37:40-80.

Self, S., M.R. Rampino and J.J. Barbera. 1981. The possible effects of large 19th and 20th
century volcanic eruptions on zonal and hemispheric surface temperatures. Journal of
Volcanology and Geothermal Research 1 1:41-60.

Sigurdsson, H. 1982. Volcanic pollution and climate: the 1783 Laki eruption. EOS 63:601-
602.

Simkin, T. and R.S. Fiske. 1983. Krakatau 1883, the Volcanic Eruption and Its Effects.
Smithsonian Institution Press, Washington, D.C. 464 pp.

Simkin, T., L. Siebert, L. McClelland, D. Bridge, C. Newhall and J.H. Latter. 1981. Volcanoes
of the World. Hutchinson Ross, Stroudsburg. 240 pp.

Stothers, R.B., J. A. Wolff, S. Self and M.R. Rampino. 1986. Basaltic fissure eruptions,
plume heights, and atmospheric aerosols. Geophysical Research Letters 13:725-728.

Sutherland, D. 1981. Weather and the peasantry of Upper Brittany. In: Climate and History.
T.M.L. Wigley, M.J. Ingram and G. Farmer (eds.). Cambridge University Press,
Cambridge, pp. 434-449.

Thorarinsson, S. 1963. Askja on Fire. Almenna Bokafelagid, Reykjavik. 48 pp.

. 1967. Surtsey, The New Island in the North Atlantic. Viking Press, New

York. 47 pp.

. 1969. The Lakagfgar eruption of 1783. Bulletin volcanologique 38:910-929.

Thordarsson, Th., S. Self, G. Larson, and S. Steinthorsson. 1987. Eruption sequence of the
Skaftar Fires 1783-1785, Iceland. EOS 68:1550.

Thordarsson, Th. and S. Self. 1988. Old maps help interpret Icelandic eruption. EOS 69:86.

Thoroddsen, Th. 1925. Die Geschichte der islandischen Vulkane. Kongelige Danske
Videnskabernes Selskabs Skrifter, Naturvidenskabelig og Mathematisk. Afdeling 8, IX.

Wang, S.W. and Z.C. Zhao. 1981. Droughts and floods in China, 1470-1979. In: Climate and
History. T.M.L. Wigley, M.J. Ingram and G. Farmer (eds.). Cambridge University
Press, Cambridge, pp. 271-288.

White, G. 1977. The Natural History of Selbourne. Penguin, New York (reprint of original
1788-89 edition).

Wood, C.A. 1984a. Amazing and portentous summer of 1783. EOS 65:409.

. 1984b. Asama 1783: lost in the rush to remember Krakatau. Volcano News 16:5.

Woodhouse, C. 1988. Tree-ring chronologies in the Great Salt Lake Basin and
paleoclimatic implications. Proceedings of the Annual Meeting of the Association of
American Geographers, (in press).

The Effects of Major Volcanic Eruptions on Canadian Surface
Temperatures

Walter R. Skinner1
Abstract

The superposed epoch method of analysis was used to detect changes in Canadian surface
temperatures due to large volcanic dust veils in the atmosphere. This method accentuates weak
signals that are present in a data series, as a temperature signal caused by a volcanic dust veil is
expected to be of the same magnitude as the background noise level. Lamb's Dust Veil Index
(DVI), a measure of the amount of volcanic material injected into the atmosphere, was used to
select the volcanic-eruption dates beginning with the eruption of Krakatau in 1883. The DVI is
directly related to the total loss of solar radiation reaching the Earth's surface, and has the
advantage of not being calculated from temperature information. Surface-temperature records for
up to 20 Canadian stations were analyzed on national, regional (Arctic) and seasonal (summer
and winter) bases for both equatorial and mid-latitude eruptions. A small sample test of
significance was applied, and all suspected temperature signals proved to be significant at the
0.01 level or better. The annual temperature depression following a mid-latitude eruption was
about 0.4°C, and occurred during the eruption year and lasted no longer. A decline in annual
surface temperature of about 1.0°C occurred in the first year after an equatorial eruption and
persisted to a lesser degree, for another year or so. The difference appears to be directly related
to the substantially greater mean DVI for the equatorial eruptions. The annual temperature drop
in the Arctic was slightly greater than that for the country as a whole. Summer-temperature
signals were stronger than those in winter, and in almost all cases were of a greater magnitude
than the annual signals. There was a marked drop in winter temperatures of about 1.0°C
following an equatorial eruption.

Introduction

The eruption of El Chichtfn in southern Mexico between 28 March and 4 April 1982 ejected huge
concentrations of gases and particles into the upper atmosphere. By mid-November 1982, detailed
solar radiation measurements in Fairbanks, Alaska began to display distinct differences from the
previous five-year normal (Wendler 1984). Clear days during the 15 November 1982 to 31 May
1983 period, when compared to clear -day data for the previous five years, showed a decrease in
the direct beam of almost 25% and a decrease in global radiation of about 5%. Mass and
Schneider (1977) previously determined that large volcanic dust veils in the atmosphere can
reduce direct solar radiation by as much as 10%. This is simultaneously accompanied by an
increased scattering effect that could substantially change the total amount of solar radiation
reaching the Earth's surface.

Many theoretical investigations (Schneider and Mass 1975; Pollack et al. 1976) and empirical
studies (Lamb 1970; Oliver 1976; Mass and Schneider 1977; Taylor et al. 1980) have been made
in an attempt to determine the possible influence of large volcanic dust veils on surface weather

1 Canadian Climate Centre, Atmospheric Environment Service, 4905 Dufferin Street, Downsview, Ontario M3H 5T4,
Canada.

and climate. Most of these investigations were conducted on either a global or hemispheric scale.
Taylor et al. (1980) also searched for volcanic signals on latitudinal, continental/marine and
seasonal bases. A drop in annual average surface temperature of between 0.5 and 1.0°C in the
first or second year following a large volcanic eruption was found in most of the empirical
studies.

Canada, having an extensive area in mid- and high latitudes, should experience volcanic dust veil
influences at varying times after a major eruption depending upon both the location and the time
of year of the eruption. Oliver (1976) estimated a mid-latitude eruption to have a same year
impact on northern hemisphere mean temperatures, while a similar impact by an equatorial
eruption would be delayed for about a year. Lamb (1970) states that the transfer of upper-level
dust veils from equatorial to mid-latitudes is accomplished mainly in autumn, and to a lesser
extent in spring, with the great seasonal circulation changes.

In this investigation, surface-temperature records for up to 20 Canadian stations were analyzed
on national, regional (Arctic) and seasonal (summer and winter) bases for both equatorial and
mid-latitude eruptions.

Methodology and Data

The Superposed Epoch Method

The superposed epoch method of analysis, as outlined by Panofsky and Brier (1965) and
employed by Mass and Schneider (1977) and Taylor et al. (1980), was used to detect changes
in Canadian surface temperatures due to large volcanic dust veils in the atmosphere. This method
accentuates weak signals that are present in a data series. A temperature signal caused by a
volcanic dust veil is expected to be of the same magnitude as the background noise level, or the
variability of the atmosphere (Taylor et al. 1980).

Volcanic Eruption Dates

Volcanic eruptions were selected on the basis of amounts of material ejected into the atmosphere,
latitude of the eruption and the isolation in time of the eruption from any other major volcanic
event. Lamb's (1970) Dust Veil Index (DVI) was used as a basis for selecting most of the
eruption dates. This is a measure of the amount of volcanic material injected into the atmosphere,
and is directly related to the loss of solar radiation reaching the Earth's surface. The eruption of
Krakatau in 1883 was given a value of 1000, and all other eruptions were adjusted to it. The DVI
has the advantage of not being calculated from temperature information (Mass and Schneider
1977).

Five of the six volcanic eruptions selected have a total DVI greater than 150, and are classed as
major volcanic events (Table 1). The 1956 eruption was chosen because of its mid-latitude
location and isolation in time from any other major volcanic event. Three equatorial and three
mid-latitude eruptions were selected in an attempt to isolate both the temporal dimensions and the
magnitudes of the temperature signals in Canada following a major volcanic event.

Selected volcanic events had to be separated by at least five years from any other major volcanic
event. This was done to avoid the problem of cumulative dust veils that might obscure resulting
signals. The separation of the 1907 event from the preceding event (1902) and the following event
(1912) is exactly five years. Incorporating the 1907 event was called for as it was a major
mid-latitude event for which ample records were available.

Table 1: Date, Location and Dust Veil Index (DVI) of Selected Major Volcanic Eruptions.

Eruption

Location

Dates Key Dates

DVI

Total

1 'Hal

DVI

Krakatau Indonesia

6.0° S

105.0° E

Aui? 1883 Aui? 1883

1000

Mont Pelftp Martininiif*

l> 1 V 'III X vIVV, 1*1 til IIIIIUUV^

15.0° N

61.0° W

Mav 1902

100

Soufriere St. Vincent

13.5° N

61.0° W

Mav 1902

300

Santa Maria, Guatemala

14.5° N

92.0° W

Oct 1902

600

Cumulative Data:

May 1902

1000

Shytubelya Sopka,

Kamchatka

52.0° N

157.5° W

Mar 1907 Mar 1907

500

Katmai, Alaska

58.0° N

155.0° W

Jun 1912 Jun 1912

150

Bezymjannaja,

Kamchatka

56.0° N

160.5° E

Mar 1956 Mar 1956

Gunung Agung, Bali

8.5° S

115.5° E

Mar 1963 Mar 1963

800

Equatorial Eruptions (#1,2 and 6)

Average Total DVI =

933

Mid-Latitude Eruptions (#3,4 and 5)

Average Total DVI =

220

Eruption year key months were also determined from Lamb (1970). The key month was the
month during which the volcano entered its most explosive phase. In the case of two or more
eruptions in the same year, such as 1902, the month of the first eruption was used. Table 1
includes the key eruption date for each selected event.

Composite Key Dates and Composited Temperatures

The key volcanic eruption date was defined as the 12-month period beginning with the month
during which the eruption occurred. The use of this period results in a cleaner volcanic signal
than using the actual calendar year of the eruption (Taylor et al. 1980). This 12-month period
was termed the "eruption year", or year "0". Sequences of four preceding years, or the four
12-month periods prior to the eruption year and the four following years, or the four 12-month
periods after the eruption year, were then determined. These sequences provided the bases for
both individual and multiple composites. The five annual periods, the eruption year and the four
following years, were analyzed because a volcanic dust veil produced by a single eruption exists
for only a few years (Lamb 1970).

Average temperature values, for selected Canadian stations, were calculated for each month of
each of the 12-month periods associated with an eruption year. The resulting 12 monthly values
were then summed and averaged to yield an annual value for that particular year. Graphs based
on individual volcanic events were then plotted and studied in an attempt to define climatic

signals.

Annual temperature values for each individual eruption were then associated with the
corresponding values for all other individual eruptions. In addition, values for equatorial
eruptions were isolated and inter-associated. The same was done for mid-latitude eruption values.
These corresponding values were then summed and averaged to yield a "superposed epoch".
Graphs, based upon these multiple volcanic events, were plotted and analyzed in a comparable
manner to the analyses of the individual events.

Data

The database used consisted of mean monthly temperature values for up to 20 Canadian weather
stations over common time periods. Stations were selected on the basis of length and
completeness of record and upon location. Thirteen stations were available for the 1883 eruption
date. There were no long-term records available for this date west of Winnipeg. Four stations
were added for the 1902 eruption date to provide east to west coast spatial coverage. Another
station was added for the 1907 and 1912 eruptions. The lack of long-term records for stations in
northern Canada restricts the study of the first four eruptions to more southerly Canadian
latitudes. Northern stations were added for the last two eruptions. This brought the total to 20
stations for the 1963 event. Table 2 shows the stations used for the 1883 eruption. Table 3 shows
the stations added for the 1902, 1907 and 1912 eruptions. Table 4 lists the stations used for the
1956 and 1963 eruptions. In some cases, such as Quebec City and Winnipeg, weather-observation
sites were moved during the 1940s from city to airport locations. However, none of the eruptions
used in this study occurred during this period. In addition, some long-term temperature records,
such as those from Toronto and Montreal, have been subjected to an artificial warming due to
the influence of urban expansion. It was hoped that this would have only minor influence on the
results and that the method of analysis would subdue the apparent noise in this small portion of
the data.

Missing monthly values were estimated for each station by calculating the 30-year mean for that
particular month. In most cases only one of the 13 to 20 values was absent. The resulting
estimate had little effect on the overall monthly composite. There were never more than two
missing values in any monthly composite.

Canadian Analysis

Taylor et al. (1980) found it necessary to use data from a group of stations rather than just
individual stations when searching for a temperature signal related to a volcanic eruption. This
is due to the year-to-year and station-to-station variability when dealing with single station
superpositions. Thus, the superposed epoch method outlined previously was applied to some or
all of the 20-station temperature database selected for this study.

Individual Eruption Composites

Figures 1 to 6 show the individual eruption dust veil temperature composites for the selected
Canadian stations. The 1883, 1902, 1956 and 1963 composites each display a marked dip in
average annual temperature either in the eruption year or in the following two years. The 1907
and 1912 composites show no such dip during these years. The low values for the "-4" and "-3"
years for the 1907 composite might be the result of a large 1902 dust veil. However, there is no
such dip in the early years of the 1912 composite that might similarly be attributed to a 1907 dust
veil.

Graphs, based on the multiple volcanic events, were then plotted and examined in an attempt to
define volcanic signals.

Table 2: Weather Stations Used in Studying the Influence of Volcanic Dust Veils on Canadian Surface
Temperatures for the 1883 Krakatau Eruption.

Weather Station

Location

Period

Years

AES No.

1. Winnipeg, Manitoba

49° 53' N,

97° 07' W

1872-1938

5023243

2. Port Arthur, Ontario

48° 26' N,

89° 13' W

1877-1941

6046588

3. Ottawa, Ontario

45° 24' N,

75° 43' W

1872-1935

6105887

4. Beatrice, Ontario

45° 08* N,

79° 23' W

1876-1979

104

6110605

5. Woodstock, Ontario

43° 07' N,

80° 45' W

1870-1981

112

6149625

6. Toronto, Ontario

43° 40' N,

79° 24' W

1840-1981

142

6158350

7. Quebec City, Quebec

46° 48* N,

71° 13' W

1872-1959

7016280

8. Montreal, Quebec

45° 30' N,

73° 35' W

1871-1981

111

7025280

9. Chatham, New Brunswick

47° 03' N,

65° 29' W

1873-1947

8100990

10. Fredericton, New Brunswick

45° 57' N,

66° 36' W

1871-1952

8101700

11. Halifax, Nova Scotia

44° 39' N,

63° 36' W

1871-1933

8202198

12. Sydney, Nova Scotia

46° 09' N,

60° 12' W

1870-1941

8205698

13. St. John's, Newfoundland

47° 34' N,

52° 42' W

1874-1956

8403500

Table 3: Weather Stations Added to Those Used for the 1883 Krakatau Eruption for the 1902, 1907 and
1912 Eruptions.

Weather Station1

Location

Period

Years

AES No.

Victoria, British Columbia

48° 25' N,

123° 22' W

1898-1981

1018610

Medicine Hat, Alberta

50° or N,

110° 43' W

1883-1981

3034480

Banff, Alberta

51° 11' N,

115° 34' W

1887-1981

3050520

Regina, Saskatchewan

50° 26' N,

104° 40, W

1883-1981

4016560

Ottawa (CDA), Ontario

45° 23' N,

75° 43' W

1889-1981

6105976

1 Ottawa (6105887) not used for 1902 eruption.

Table 4: Weather Stations Used In Studying the Influence of Volcanic Dust Veils on Canadian Surface
Temperatures for the 1956 Bezymjannaja and 1983 Gunung Agung Eruptions.

Years
Data of

Weather Station1 Location Period Record AES No.

1. Victoria, British Columbia 48° 25' N, 123° 22' W 1898-1981 84 1018610

2. Medicine Hat, Alberta 50° 01' N, 110° 43' W 1883-1981 99 3034480

3. Banff, Alberta 51° 11' N, 115° 34' W 1887-1981 95 3050520

4. Regina, Saskatchewan 50° 26' N, 104° 40' W 1883-1981 99 4016560

5. Winnipeg, Manitoba 49° 53' N, 97° 07' W 1938-1981 44 5023243

6. Churchill, Manitoba 58° 45' N, 94° 04' W 1943-1981 39 5060600

7. Ottawa CDA, Ontario 45° 23' N, 75° 43' W 1889-1981 93 6105976

8. Beatrice, Ontario 45° 08' N, 79° 23' W 1876-1979 104 6110605
9 Woodstock, Ontario 43° 07' N, 80° 45' W 1870-1981 112 6149625

10. Toronto, Ontario 43° 40' N, 79° 24' W 1840-1981 142 6158350

11. Quebec City A, Quebec 46° 48' N, 71° 23' W 1843-1981 39 7016294
12 Montreal, Quebec 45° 30' N, 73° 35' W 1871-1981 111 7025280

13. Chatham A, New Brunswick 47° 01' N, 65° 27' W 1943-1981 39 8101000

14. Fredericton CDA, N.B. 45° 55' N, 66° 37' W 1913-1981 69 8101600

15. Halifax, Nova Scotia 44° 39' N, 63° 34' W 1939-1974 36 8202200

16. Sydney A, Nova Scotia 46° 10' N, 60° 03' W 1941-1981 41 8205700

17. St. John's, Newfoundland 47° 35* N, 52° 44' W 1957-1975 19 8403501

18. Cambridge Bay, N.W.T. 69° 07' N, 105° 01' W 1929-1981 53 2500600

19. Mould Bay, N.W.T. 76° 14' N, 119° 20' W 1948-1981 34 2502700

20. Kuujjuaq, Quebec 58° 06' N, 68° 25' W 1947-1981 35 7112400

1 St. John's (8403501) not used for the 1956 eruption.

KRAKATAU (1883)

KATNAI (1912)

YEARS FROM ERUPTION YEAR

Figure 1: Dust veil temperature composite.

Thirteen Canadian station events.

YEARS FROM ERUPTION YEAR

Figure 4: Dust veil temperature composite.

Eighteen Canadian station events.

THREE ERUPTIONS (1902)

BEZYMJANNAJA (1956)

G 6-i
r 5.9

^ 5.7

5 . 5
W 5.3
« 5.1

g 4.7
K 4.5
W 4.3
4.1
* 3.9

3 . 7

-4-3-2-10 1 2 3 4

YEARS FROM ERUPTION YEAR

Figure 2: Dust veil temperature composite.

Seventeen Canadian station events.

YEARS FROM ERUPTION YEAR

Figure 5: Dust veil temperature composite.

Nineteen Canadian station events.

SHYTUBELYA (1907)

GUNUNG AGUNG (1963)

YEARS FROM ERUPTION YEAR

Figure 3: Dust veil temperature composite.

Eighteen Canadian station events.

j IlllLli

H 1.9
1 .?

r 1.3
w

-4 -3 -2 -1

12 3 4

YEARS FROM ERUPTION YEAR

Figure 6: Dust veil temperature composite.

Twenty Canadian station events.

The apparent significance of these graphs must be viewed with caution. The first four eruptions
were embedded in a hemispheric-warming trend, whereas the last two eruptions occurred during
a hemispheric-cooling trend (Mass and Schneider 1977). The year-to-year variability, or noise,
found by Taylor et al. (1980) is evident in a Canadian context. The compositing of several
volcanic events should reduce this noise level and accentuate a volcanic dust veil signal.

Multiple Eruption Composites

Figure 7 shows the temperature composite for all stations and all eruptions. There is an obvious
temperature dip during the eruption year and the " + 1" year. The temperature dip during these
two years is about 0.4°C below the level of years "-4" to "-1". Figure 8 shows the composite
for the three equatorial eruptions. There is a well-marked dip in the " + year, about 1.1 °C
beiow the level of years "-4 to "-1". Figure 9 shows the composite for the three mid-latitude
eruptions. Here the temperature dip is in the actual eruption year, about 0.5° C below the levels
of years "-4" to "-1".

Arctic Analysis

The data noise level, or year-to-year variability, when based upon different groupings of stations,
should vary randomly while the volcanic signal should remain fairly constant (Taylor et al.
1980). A regional analysis was one step in determining the significance of the possible volcanic
signals outlined previously. It also provided the basis for volcanic signal investigation into a part
of Canada which can be extremely sensitive to small alterations in surface temperature.

The solar-radiation deficit produced by volcanic dust veils must be greatest in Arctic areas where
dust veils persist longer and the sun's rays travel obliquely through the layers of dust (Lamb
1970). Reduced surface temperatures result in an accumulation of both sea ice and land snow.
The increased albedo would produce a radiation deficit long after the dust veil has disappeared
(Lamb 1970). It would also affect the general atmospheric circulation, possibly having
far-reaching spatial effects.

The superposed epoch analysis method was applied to four Canadian Arctic stations for the 1956
and 1963 eruptions. There were no Canadian Arctic station records for the earlier eruptions. The
stations used were Churchill, Manitoba, Cambridge Bay, Northwest Territories, Mould Bay,
Northwest Territories and Kuujjuaq, Quebec.

Figure 10 shows the average annual temperature composite for the 1956 and 1963 eruptions. The
temperature dip in the "0" and " + 1" years is similar to that in Figure 7 for all Canadian stations.
It is about 1°C below the levels of years "-4" to "-1". The surrounding noise level, however, is
quite different than that in Figure 7. Years " +2" to "+4" hint at Arctic temperature stability
following a volcanic eruption.

Seasonal Analyses

Summer and winter investigations were made in an attempt to determine the relative magnitudes
of the dust-veil signals. The key summer season was defined as the first three-month period (June
to August) to follow an eruption. The key winter season was defined as the first three-month
period (December to February) to follow an eruption. Sequences of four preceding and four
following seasons were determined in the same manner outlined previously. Seasonal averages
were calculated for all Canadian stations and for each year associated with a volcanic eruption.

K
H

K
W
CL,
Z

ALL ERUPTIONS

-4 -3 -2 -1

1 2 3

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 7: Dust veil temperature composite.
All Canadian station events.

EQUATORIAL ERUPTIONS

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 8: Dust veil temperature composite.

Fifty Canadian station events for
three equatorial eruptions.

MID-LATITUDE ERUPTIONS

4. 7n
~ 4.5
g 4.3-

g 3.7^
z 3.5

W -4-3-2-10123

llljll,

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 9: Dust veil temperature composite.

Fifty-four Canadian station events
for three mid-latitude eruptions.

1956 & 1963 ERUPTIONS

Summer

Figures 11 to 13 show the summer season temperature composites. All but one graph show a
distinct drop of up to several tenths of a degree in either the eruption year or the following year.
These composites display a close resemblance to those in Figures 7 to 9. The magnitude of each
temperature drop, however, is at least equal to or greater than that of the corresponding annual
composite.

Winter

Figures 14 to 16 show the winter season temperature composites. There is a higher degree of
year-to-year variability than there was in the summer composites. This makes it more difficult
to detect a possible volcanic signal. There is a distinct drop in temperature during the first winter
following an equatorial eruption (Figure 15). However, there is no such drop in temperature
following a mid-latitude eruption (Figure 16).

Significance Tests

The fact that the regional Arctic analysis identified much the same volcanic signals as those of
the national study is a supportive indication of significance. A more rigorous small-sample test,
however, is desirable.

The Student t-test was applied to the multiple eruption composites to determine whether the
sample mean, or the mean of the one or two years during which the volcanic signal is evident,
is significantly different than the population mean, or the mean of the nine years from which it
was taken. Mass and Schneider (1977) applied this test to volcanic dust veil composites for
northern hemisphere stations. The basic formula used was:

t= (~x-M)^N

where, x = sample mean

\i = population mean

a = population standard deviation
and, N = average number of stations, and x = number of eruptions in the composite.

ALL ERUPTIONS

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 11: Dust veil summer season temperature
composite. All Canadian station
events.

EQUATORIAL ERUPTIONS

. 1

. 9

. 7

. 5

3
H

. 3

. 1

PER

. 9

. 7

-4 -3 -2-10 1

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 12: Dust veil summer season temperature
composite. All Canadian station
events.

MID-LATITUDE ERUPTIONS

-4-3-2-19 1 2 3

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 13: Dust veil summer season temperature
composite. All Canadian station
events.

ALL ERUPTIONS

K
=J
H

K
W

-4 -3 -2-10 1

RS BEFORE OR AFTER

2 3 4

ERUPTION YEAR

Figure 14: Dust veil winter season temperature
composite. All Canadian station
events.

EQUATORIAL ERUPTIONS

U
e

W
K
a

H
<t
K
W

-4-3-2-10 1 2 3 4

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 15: Dust veil winter season temperature
composite. All Canadian station
events.

MID-LATITUDE ERUPTIONS

V -8.5 n

H -4-3-2-101234

YEARS BEFORE OR AFTER ERUPTION YEAR

Figure 16: Dust veil winter season temperature
composite. All Canadian station
events.

Thus, for a mid-latitude composite for all stations

N = 18 x 3 = 54

The degrees of freedom (Gregory 1963) are

d.f. = (n-1) + (n-1)
= n, + n,, - 2

where, r^ = number of population years

nb = number of sample years
d.f. (1 sample year) = 8
d.f. (2 sample years) = 9

Table 5: Student t-Test Calculations for Composites Having an Apparent Volcanic Signal.

Figure Composite x Years /x a N d.f. t a

I All Stations

All Events 3.70 (0,1) 3.91 0.21 108 9 10.4 0.001

8 All stations

Equatorial Events 2.91 (1) 3.71 0.37 51 8 15.2 0.001

9 All Stations

Mid-Latitude Events 3.68 (0) 4.13 0.25 54 8 13.2 0.001

10 Arctic (a) -11.97 (0,1) -11.34 0.44 8 9 4.1 0.01

(b) -12.15 (0) - - - 8 5.2 0.001

I I Summer
All Stations

All Events 16.09 (0,1) 16.57 0.33 108 9 15.1 0.001

12 Summer
All Stations

Equatorial Events 15.93 (1,2) 16.48 0.41 51 9 9.6 0.001

13 Summer
All Stations

Mid-Latitude Events 15.95 (0) 16.64 0.39 54 8 13.0 0.001

15 Winter

All Stations

Equatorial Events -10.63 (1,2) -9.76 0.72 51 9 8.6 0.001

The problem encountered earlier concerning the low number of stations and eruptions when
dealing with the Arctic composites needs to be discussed. The fewer the stations and eruptions
used, the greater the difference between the means must be, in order to attain a given level of
significance. Table 5 shows the calculated Student-t values and the associated levels of
significance for all composites where a volcanic signal was apparent. The test results are similar
to those of Mass and Schneider (1977). In all cases, there is a difference between the sample
population and the entire set at a significance level («) of at least 0.01.

Conclusions

Climatic variation is complex and influenced by many factors. It is therefore difficult to clearly
identify possible volcanic influences. As a result, caution must be exercised when interpreting
apparent historical evidence and using it to predict future events. However, the results of this
study do provide some evidence of the effects of volcanic dust veils on surface temperatures. This
allows some tentative conclusions to be made.

The magnitude of the annual temperature drop, for all Canadian stations, was at least 0.5 °C
greater after the equatorial eruptions analyzed than after the mid-latitude eruptions analyzed. The
average total DVI for the selected equatorial eruptions was 933, while it was 220 for the
mid-latitude events. The mid-latitude temperature depression was about 0.4°C, occurring during
the eruption year and lasting no longer. The equatorial signal of about 1 .0°C occurred in the first
year after the eruption year and persisted to a lesser degree, for another year or so.

The annual temperature drop in the Arctic was slightly greater than that for the country as a
whole. It was approximately 1.0°C, and occurred in both the eruption year and the year
following. The lower significance levels for the Arctic signals reflects the small number of
stations and events used. Further investigation of this region, using more stations and events,
might be appropriate.

Temperature signals were stronger in the summer than in the winter. In addition, the summer
drops in temperature were, in almost all cases, of a greater magnitude than the annual drops.
There was a marked drop in winter temperature of about 1.0°C in the year following an
equatorial eruption.

This investigation did not take trends of temperature into account. No technique, other than the
compositing of several volcanic events, was used to eliminate trends. The first four eruptions
selected occurred during hemispheric-warming trends, whereas the latter two occurred during
hemispheric-cooling trends. An accurate assessment of volcanic dust veil signals would eliminate
these trends before applying the compositing technique. The temperature results found in this
investigation are quantitatively similar to the empirical results found by Mass and Schneider
(1977) and Taylor et al. (1980) and to the theoretical results of Pollack et al. (1976).

Acknowledgements

This project was undertaken in the Applications and Impact Division of the Canadian Climate
Centre, Environment Canada. Mr. M.O. Berry provided project supervision.

References

Gregory, S. 1963. Statistical Methods and the Geographer. Longmans, Green and Co. Ltd.,
London. 240 pp.

Lamb, H.H. 1970. Volcanic dust in the atmosphere; with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society, London
266:425-533.

Mass, C. and S.H. Schneider. 1977. Statistical evidence on the influence of sunspots and volcanic
dust on long-term temperature records. Journal of Atmospheric Science 34:1995-2004.

Oliver, R.C. 1976. On the response of hemispheric mean temperature to stratospheric dust: an
empirical approach. Journal of Applied Meteorology 15:933-950.

Panofsky, H.A. and G.W. Brier. 1965. Some Applications of Statistics to Meteorology. First
Edition. Pennsylvania State University, pp. 159-161.

Pollack, J.B., O.B. Toon, C. Sagan, A Summers, B. Baldwin and W. Van Camp. 1976.
Volcanic eruptions and climatic change: a theoretical assessment. Journal of Geophysical
Research 81:1071-1083.

Schneider, S.H. and C. Mass. 1975. Volcanic dust, sunspots and temperature trends. Science
190:741-746.

Taylor, B.L., T. Gal-Chen and S.H. Schneider. 1980. Volcanic eruptions and long-term
temperature records: an empirical search for cause and effect. Quarterly Journal of the
Royal Meteorological Society 106:175-199.

Wendler, G. 1984. Effects of the El Chichdn volcanic cloud on solar radiation received at
Fairbanks, Alaska. Bulletin of the American Meteorological Society 65:216-218.

Northern Hemisphere

North America

Climate of 1816 and 1811-20 as Reconstructed from Western North
American Tree-Ring Chronologies

J.M. Lough1
Abstract

Reconstructed temperature and sea-level pressure anomalies are presented for the year 1816 and
the decade 181 1-20. The reconstructions were developed from western North American semi-arid
site tree-ring chronologies. The reconstructed climatic conditions for North America and the
North Pacific were not very anomalous for either 1816 or 181 1-20. More unusual conditions were
reconstructed in years other than 1816 between 1811-20, and for decades other than 1811-20 in
the first half of the nineteenth century. The factors responsible for the unusual climatic conditions
of the "year without a summer" do not appear to have affected surface climate of western North
America to the extent that these conditions are translated into the climatic reconstructions.

Introduction

The exceptionally large eruption of Tambora in April 1815 has frequently been speculated to have
been the cause of the unusual climatic conditions experienced in 1816 - "the year without a
summer". Anomalous weather was recorded in that year in eastern North America and Europe
(Milham 1924; Rampino and Self 1982; Stommel and Stommel 1983; Stothers 1984; and
elsewhere in this volume). The extent of climatic anomalies outside of the regions bordering the
North Atlantic has not, as yet, been appraised satisfactorily.

Although empirical studies have provided evidence of large-scale area-averaged surface
temperature decreases following major volcanic eruptions (e.g., Oliver 1976; Taylor et al. 1980;
Self et al. 1981; Kelly and Sear 1984; Sear et al. 1987) and the results of a variety of models
have supported the role of volcanic eruptions as a source of thermal forcing (e.g., Hunt 1977;
Robock 1981; Gilliland 1982; Gilliland and Schneider 1984), the importance of volcanic eruptions
(such as Tambora in 1815) as a major source of climatic variability is still disputed (e.g.,
Deirmendjian 1973; Landsberg and Albert 1974; Parker 1985; Ellsaesser 1986). Difficulties in
assessing the role of volcanic eruptions in climatic variability arise for a number of reasons.
Theoretical (e.g., Baldwin et al. 1976; Pollack et al. 1976) and empirical studies (e.g., Rampino
and Self 1982, 1984) indicate that the amount of sulphate aerosols produced by an eruption is of
more importance than the amount of silicate ash in determining the subsequent climatic impact.
Unfortunately, most historical chronologies of volcanic eruptions (e.g., Lamb 1970; Hirschboeck
1979-80; Newhall and Self 1982) do not provide measures of sulphate aerosols, only of the
explosive magnitude of the eruptions, which is often assessed by the amount of ash produced.
Acidity profiles from ice cores (e.g., Hammer et al. 1980; Legrand and Delmas 1987) can
provide records of eruptions that produced considerable amounts of sulphuric acid aerosols. The
ice-core records tend, however, to be biased towards eruptions occurring at higher latitudes at
the expense of those occurring at lower latitudes, and so such records tend to be incomplete.

Australian Institute of Marine Science, PMB 3, Townsville M.C., Queensland 4810, Australia.

Other problems result from the small number of possibly climatically important volcanic eruptions
that have occurred during the period for which extensive instrumental climatic records are
available. The small sample size limits the statistical inferences that can be made regarding the
impact of volcanic eruptions on climate. Consequently, most empirical studies have examined
temperature series averaged over zonal or hemispheric space scales and little attention has been
given to the possible regional variations of a climatic response. For periods before the mid-
nineteenth century, instrumental records can provide information for geographically limited
regions, usually those bordering the North Atlantic (e.g., Angell and Korshover 1985). For
periods prior to the introduction of widespread instrumental climatic records we must rely on
proxy climatic information from documentary, geological and biological sources. Lough and
Fritts (1987), for example, identified a possible spatial response of North American temperatures
to low-latitude volcanic eruptions. The response comprised warming in the western states and
cooling in the central and eastern states. This study was based on temporally and spatially detailed
reconstructions of North American temperatures derived from western North American tree-ring
chronologies, and covered the period from 1602 to 1900 A.D. Some verification of the
reconstructed climatic response was provided by independent sources of proxy climatic
information both within and outside of the study area. This is important as each proxy climatic
record is an imperfect record of past climate. Each series contains bias and error terms which
may be unrelated to climate. In addition, different series may respond to different climatic
variables, in different seasons and with different frequency responses. The most comprehensive
description and understanding of past climatic variations (and their possible causes) will,
therefore, only be obtained by the careful comparison and integration of independent sources of
information (e.g., National Academy of Science 1975; National Science Foundation 1987).

As a contribution to the improved description and understanding of the climate of 1816 and the
decade 181 1-20, I present reconstructions of seasonal climate for North America and the North
Pacific developed from western North American tree-ring width chronologies.

Data

The reconstructions used in this study were developed by H.C. Fritts and co-workers at the
Laboratory of Tree-Ring Research, Tucson, Arizona, following the methods outlined by Fritts
et al. (1979) and described in detail by Fritts (in press). Only a general description of some of
the characteristics of these reconstructions is given here. Fritts (1976), Hughes et al. (1982) and
Stockton et al. (1985) describe the general principles and procedures applied in
dendroclimatology.

An array of 65 low-altitude, semi-arid site tree-ring chronologies (Figure 1; Fritts and Shatz
1975) was used to estimate, by canonical regression, seasonal values of temperature at 77
stations, precipitation at 96 stations in the United States and southwestern Canada, sea-level
pressure at 96 stations in the United States and southwestern Canada and sea-level pressure at 96
gridpoints between 100°E and 80°W, 20°N and 70°N. Because of the general west to east
movement of weather systems across North America it was possible to attempt reconstruction of
climate outside the area covered by the tree-ring predictor grid (see also Kutzbach and Guetter
1980). The temperature and precipitation models were calibrated over the period 1901-63, and
those for sea-level pressure from 1899-1963. The temperature and precipitation estimates were
verified with data independent of that used for model calibration. The general form of the final
sea-level pressure models was verified using a subsample replication technique (Gordon 1982).

The final reconstructions, representing the average of the two or three best-calibrated and verified
models, were for each variable, station or gridpoint and season for the years from 1602 to 1961.
The seasons were December to February (DJF), March to June (MAMJ), July to August (JA)
and September to November (SON). The annual series were the average of the four seasonal
reconstructions and, therefore, were from December to November. In retrospect, the use of the
four- and two-month seasons has proved a drawback in comparing these reconstructions to other
sources of information.

The annual calibration and verification statistics can provide some insight into the reliability of
these reconstructions. More than 30% of the temperature variance was explained over North
America with values exceeding 50% over much of the central United States (Figure 2). Most of
the region also showed reliability through positive reduction of error (RE) statistics and the
majority of verification tests passed. Positive values of RE indicate that, over the verification
period, the estimates are an improvement over simply assuming mean climatic conditions (Gordon
1982). Areas of poor temperature reliability occurred in the northeastern United States, Florida
and parts of Nevada and Colorado.

Generally, less variance was calibrated for precipitation than temperature (Figure 2), with more
than 30% variance explained only in an area extending along the eastern edge of the tree-ring
predictor grid. Verification of the precipitation estimates was also poor over much of the region.
The precipitation reconstructions appeared to be of much lower reliability throughout much of
the southern and eastern United States. The canonical regression transfer function (which is based
on matching of the large-scale patterns of climate and tree-growth represented by the major
principal components of the respective grids) does not appear to be well-suited to the
reconstruction of precipitation. This variable is dominated by small-scale processes and variability
that are not well captured by this regression technique. This is despite the fact that the tree-ring
chronologies used are most directly sensitive to precipitation (Fritts 1974).

The calibration and verification statistics for sea-level pressure (Figure 3) show that more than
30% of the variance was calibrated over a large part of the grid. The reconstructions tended to
be least reliable over northeastern Asia - the area farthest removed from the tree-ring predictor
sites.

Other general features of these reconstructions were (Fritts, in press): (a) sea-level pressure
tended to be biased towards lower frequency climatic variations at the expense of high frequency,
year-to-year variations; (b) autumn climate was poorly reconstructed for all three variables;
(c) precipitation was least well reconstructed, and temperature was probably the most reliably
reconstructed; (d) all reconstructions deteriorated in reliability downstream from the tree-ring
predictor grid over eastern North America (where Atlantic influences outweigh those of the
Pacific) and, for sea-level pressure, over eastern Asia; (e) the large-scale regional patterns of
climatic variation were calibrated at the expense of precision at individual stations or gridpoints;
and (0 the reliability of the reconstructions was enhanced by averaging over space and filtering
through time.

A full description of these reconstructions and their development is provided by Fritts (in press).
Trie reconstructions have been applied in a number of studies into the nature of climatic variations
in North America and the North Pacific and also compared with independent sources of climatic
information (e.g., Fritts and Lough 1985; Gordon et al. 1985; Lough and Fritts 1985, 1987;
Lough et al. 1987). These studies, together with analyses of the reconstructions themselves (Fritts
in press), have provided insights into the strengths and weaknesses of this particular set of
climatic estimates. In the words of H.C. Fritts (in press): "The specific conclusions regarding the
climate from 1602 to 1960 are presented as tentative hypotheses derived from one dendroclimatic
analysis and test. They must be compared to data from other independent paleoclimatic sources
that can reveal changes on seasonal and decadal time scales with accurate yearly dates".

CALIBRATED VARIANCE

VERIFICATION STATISTICS

ANNUAL PRECIPITATION

Figure 2: Calibration and verification statistics for annual temperature (top) and annual precipitation
(bottom). Percent variance explained with areas of greater than 30% shaded (left-hand
figures); number of verification tests passed out of a total of five, and areas with RE statistic
greater than zero shaded (right-hand figures).

101

CALIBRATED VARIANCE

100E 120 140 160 180 160 140 120 100 80W

VERIFICATION STATISTICS

100E 120 140 160 180 160 140 120 100 80W

ANNUAL PRESSURE

Figure 3: Calibration and verification statistics for annual sea-level pressure. Notation as for Figure 2.

102

Results

The seasonal and annual reconstructed values of temperature and sea-level pressure for 1816 and
1811-20 were compared with the reconstructed mean climate of 1901-60. The temperature
reconstructions were standardized by the 1901-60 standard deviation (s.d.). The reconstructions
and original 65 tree-ring chronology series were also compared with the mean of the whole
period, 1602 to 1960, to assess how unusual 1816 and 181 1-20 were in the longer-term context.

1816

The seasonal and annual reconstructions of temperature and sea-level pressure are presented in
Figure 4. In the winter of 1815-16, the Aleutian Low was reconstructed to be displaced
south eastwards, with slightly higher pressure reconstructed over the Canadian Arctic.
Temperatures were reconstructed to be warmer in the western states (associated with enhanced
southerly air flow) and cooler over the central and eastern states. Temperature departures up to
2 s.d. below recent-period means were reconstructed over the Great Lakes.

The large sea-level pressure anomalies reconstructed in spring over eastern Asia were in an area
of low reconstruction reliability and were not, therefore, considered to be significant. The main
reconstructed feature was a slight deepening of the Aleutian Low. The reconstructed temperature
field did not exhibit very large anomalies, though temperatures were still warmer in the west and
cooler in the central states compared to the 1901-60 normals.

Discounting the sea-level pressure anomalies over eastern Asia, the reconstructed sea-level
pressure field for summer did not show marked departures from the twentieth century mean
values. Slightly lower pressure was reconstructed over western Hudson Bay. Temperatures were
reconstructed to be slightly above the average over a large part of the United States, with below
average conditions in the far western states. Temperatures were reconstructed to be close to the
1901-60 mean over the northeastern United States, the area of extensively documented climatic
anomalies for the summer of 1816.

In autumn, a positive pressure anomaly was reconstructed over the eastern North Pacific that was
linked with the colder temperatures reconstructed in the Pacific Northwest. Elsewhere in the
United States, temperatures were reconstructed to be warmer than the 1901-60 mean values by
up to 2 s.d. in the northeastern and southern states. However, the reconstructions are least
reliable in autumn.

In the annual average, discounting sea-level pressure anomalies over Asia, the major
reconstructed feature was an area of higher pressure to the west of Hudson Bay. Higher sea-level
pressure extended out over the Pacific, and lower pressure was reconstructed to the south. Thus,
1816 seems to have been characterized by a weakened zonal circulation over the North Pacific.
Temperatures were reconstructed to be warmer in the western states and cooler in the most
southerly states. Although temperatures were reconstructed to be up to 1 s.d. below the 1901-60
mean near the Great Lakes and northeastern United States, the main contribution to this appears
to come from the temperatures reconstructed for the winter 1815-16.

1811-20

Figure 5 shows the reconstructed seasonal and annual sea-level pressure and temperature values
averaged for the decade of 181 1-20. The reconstructions were expressed as departures from the
reconstructed mean of 1901-60, and those for temperature were standardized.

103

a) DJF

b] MAMJ

e) Annual

Figure 4: Reconstructed sea-level pressure (mb) and temperatures (s.d. units) expressed as departures
from the 1901-60 means the year 1816 for: (a) winter; (b) spring; (c) summer; (d) autumn;
and (e) annual data.

104

e) Annual

Figure 5: Reconstructed sea-level pressure (mb) and temperatures (s.d. units) expressed as departures
from the 1901-60 means for the decade 1811-20 for: (a) winter; (b) spring; (c) summer; (d)
autumn; and (e) annual data.

105

In winter, the Aleutian Low was reconstructed to be deeper than the average, with positive sea-
level pressure departures over the Canadian Arctic. Temperatures were reconstructed to be cooler
than the average through the central United States.

In spring a negative sea-level pressure anomaly was reconstructed over Alaska with near-average
conditions reconstructed elsewhere. Temperatures were reconstructed to be warmer than the
average throughout most of the United States. These departures were significantly different from
the 1901-60 mean, at the 5% level, for 73% of the 77 temperature stations.

The summer sea-level pressure anomalies were reconstructed to be of small magnitude, with the
exception of northeastern Asia. Temperatures were reconstructed to be cooler than average in the
northwestern states and generally warmer than average in the central and eastern regions. There
was no evidence in these reconstructions of negative temperature anomalies in the eastern United
States.

The autumn sea-level pressure field was characterized by a positive anomaly in the northeastern
North Pacific. Temperatures were reconstructed to be cooler than the average in the northwestern
and western regions and warmer in the southeastern and eastern regions.

In the annual average, the sea-level pressure anomalies (outside of Asia) were estimated to be
of small magnitude. Slightly below average pressure was found over the North Pacific.
Temperatures were reconstructed to be slightly warmer than the average over most of the United
States, though at only 5% of the 77 stations were these values significantly different from the
1901-60 mean values.

Thus, the climate of 1811-20, as reconstructed from western North American tree-ring
chronologies, did not appear to be particularly anomalous when compared to the mean climate
of 1901-60. In the annual average, sea-level pressure was slightly lower and temperature slightly
higher than the 1901-60 mean, but none of these departures was very large.

The reconstructed climate of 1811-20 was compared with that reconstructed for the other four
decades of the first half of the nineteenth century (Figure 6). These data were expressed as
departures from the instrumental record mean of 1901-70, and precipitation was included,
expressed as a percentage of the mean. In this context, 181 1-20, appeared to have been the least
unusual of the five decades. Extensive cooling was, for example, reconstructed in 1821-30, 1831-
40 and 1841-50. Similarly, sea-level pressure anomalies of greater than 1 mb were evident in all
decades except 181 1-20. The climate as reconstructed from the western North American tree-ring
chronologies for the decade 1811-20 was not very different from the recent mean conditions.
More extreme climatic conditions were reconstructed for other decades in the first half of the
nineteenth century.

Comparisons with 1602-1960 Mean Conditions

In the preceding sections the reconstructed climate of 1816 and 181 1-20 was compared to recent,
twentieth century mean conditions. The reconstructed climate did not appear to be very different
from this mean. I examined the data with respect to the long-term 1602-1960 reconstruction
mean. I also considered the nature of the anomalies of the original tree-ring chronologies which
were used to develop these climatic reconstructions.

106

Figure 6: Reconstructed annual sea-level pressure (mb), temperatures (°C) and precipitation (percent of
mean) expressed as departures from the 1901-70 instrumental record means for the first five
decades of the nineteenth century. Dashed contour lines for the precipitation maps are through
areas where the verification statistics indicate that the reconstructions are unreliable.

107

The percentage of the 77 annual temperature stations and 65 tree-ring chronologies with
departures of +1 s.d. and -1 s.d. of the 1602-1960 mean were calculated for each year of the
decade 1811-20 (Table 1). The reconstructed temperature field was close to average conditions
with only 3% of the stations with reconstructed values ± 1 s.d. of the mean in 1816. The years
1811, 1818 and 1819 all had more than 45% of stations with departures ±1 s.d. from the mean.
In 1811, the departures were about equally above and below the mean, but in 1818 and 1819,
they were mainly positive, indicating warmer conditions.

Forty percent of the original tree-ring chronologies had departures of at least ± 1 s.d. of the
1602-1960 mean in 1816, though this was not the most extreme year of the decade. The most
extreme years were 1819 with 48% and 1818 with 42% of the 65 sites with departures ± 1 s.d.
of the mean. For the last two years, the departures were mainly negative, indicating that
conditions were generally unfavourable for tree growth. In contrast, in 1816, 35% of the 65
stations had departures of at least + 1 s.d. of the mean, indicating conditions were generally
favourable for wide growth-ring formation in western North America. This was the most
favourable year for the tree growth of the decade 181 1-20. The term favourable for tree growth
cannot be simply interpreted, as the 65 chronologies cover a range of tree species from different
sites in western North America. Factors influencing the width of the annual growth ring vary
considerably, and can also operate over a number of growing seasons (Fritts 1976). For semi-arid
sites, wider annual rings are often, however, associated with moister and cooler conditions near
the trees.

The decade mean for each reconstructed variable and the tree-ring chronologies were compared
to the long-term mean for 1602-1960 for each decade between 1602-10 (1602 was the first year
of the reconstructions) and 1951-60 (Table 2). Evidently 1811-20 was not particularly unusual
in these data. For temperature, 12% of the stations had departures significantly different from
the long-term mean in 181 1-20 compared to 62% of stations in the most extreme decade of 1681-
90. For sea-level pressure, 1811-20 had 33% of the 96 gridpoints with significant departures
compared to the most unusual decade of 1881-90 with 66%. None of the 96 precipitation stations
was reconstructed to have values significantly different from the long-term mean in 1811-20,
compared to 70% in 1611-20. For the original tree-ring chronologies, 19% were significantly
different in 1811-20, compared to 60% for the most extreme decade of 1911-20.

108

Table 1: Percentage of 77 Temperature Stations and 65 Tree-Ring Chronologies with Values ± 1 Standard
Deviation of 1602 to 1960 Mean for Each Year of the Decade 1811-20.

Annual Temperature

Year

s.d.
+ 1

1811
1812
1813
1814
1815
1816
1817
1818
1819
1820

29
27
9
0
14
3

10
47
38
26

s.d.
-1

s.d.
±1

25
10
1
8

12
0
0
0
8
7

53
37
10

8
26

10
47
46
33

Tree-Ring Chronologies

s.d. s.d. s.d.

+ 1 -1 +1

1811 15 6 22

1812 9 6 15

1813 6 23 29

1814 12 11 23

1815 14 9 23

1816 35 5 40

1817 22 5 27

1818 11 31 42

1819 14 34 48

1820 8 31 39

109

Table 2: Percentage of Stations, Gridpoints or Chronologies for Which Decade Mean is Significantly
Different from Long-Term (1602-1960) Mean at the 5% Significance Level for Reconstructed
Temperature (T), Sea-Level Pressure (SLP), Precipitation (PPT) and Tree-Ring Chronologies
(TREES).

DECADE T SLP PPT TREES

1602-1610

1611-1620

1621-1630

1631-1640

1641-1650

1651-1660

1661-1670

1671-1680

1681-1690

1691-1700

1701-1710

1711-1720

1721-1730

1731-1740

1741-1750

1751-1760

1761-1770

1771-1780

1781-1790

1791-1800

1801-1810

1811-1820

1821-1830

1831-1840

1841-1850

1851-1860

1861-1870

1871-1880

1881-1890

1891-1900

1901-1910

1911-1920

1921-1930

1931-1940

1941-1950

1951-1960

—

110

Summary and Conclusions

As reconstructed by western North American semi-arid site tree-ring chronologies, the climate
of North America and the North Pacific does not appear to have been very unusual in 1816 or
the decade 181 1-20. This is when compared to both the 1901-60 and the 1602-1960 reconstructed
data means.

For winter, spring and the annual average of 1816, temperatures were reconstructed to be cooler
in the eastern and central United States and warmer in the western United States. In summer and
autumn of 1816, temperatures were reconstructed to be warmer in the central and eastern regions
and cooler in the west. The pattern of temperature departures for winter, spring and the annual
average are similar to the average pattern identified by Lough and Fritts (1987) to characterize
the years 0 to 2 after eight low-latitude volcanic eruptions between 1602 and 1900. The Tambora
eruption of 1815 was one of eight eruptions used in that analysis. The summer temperature field
for 1816 does not resemble the average pattern identified by Lough and Fritts (1987). In the
annual average there was reconstructed to be a weakening of the westerly zonal flow pattern over
the North Pacific. Sea-level pressure anomalies were not, however large. Thirty-five percent of
the original tree-ring chronologies had growth departures of +1 s.d. or more above the 1602-
1960 mean in 1816, indicating that conditions, at least in parts of western North America were
generally favourable for wide tree-ring formation.

The decade 1811-20, in the annual average, was reconstructed to be slightly warmer than the
1901-60 mean over North America, with lower sea-level pressure reconstructed over the North
Pacific. It was, perhaps, the least unusual of the first five decades of the nineteenth century.
Relatively large negative temperature departures were reconstructed over the central northern
United States in 1821-30, 1831-40 and 1841-50. The decade 181 1-20 did not appear to be very
unusual when compared to long-term mean conditions for any of the reconstructed variables nor
the original tree-ring chronologies.

The evidence from this particular set of climatic reconstructions from western North American
semi-arid site tree-ring chronologies is for near-normal climatic conditions in 1816 and 181 1-20.
Reconstructed climatic anomalies were small in magnitude when compared to the recent, 1901-60,
and long-term 1602-1960, mean conditions. Most references to the "year without a summer" in
North America tend to come from eastern regions. Because this particular set of reconstructions
is known to be less reliable in the east, where Atlantic and Arctic influences outweigh those of
the Pacific, the lack of large reconstructed anomalies in this region was not surprising. What was
surprising was a lack of evidence for large-magnitude climatic anomalies in areas where the
reconstructions are known to be reliable, over the western United States and the North Pacific.
Analysis of the original tree-ring chronology series suggested that 1816 was a year favourable
for tree-growth in parts of the western states, possibly associated with moister and cooler
conditions. Large-scale climatic anomalies are not, however, apparent in the climatic
reconstructions from these tree-ring data. This suggests that whatever the nature of the anomalies
of climate in 1816 and the decade 181 1-20, they were not large enough to significantly influence
climatic conditions in the western United States either for good or bad.

Acknowledgements

This study is based on the results of many years of work by Hal Fritts and co-workers at the
Laboratory of Tree-Ring Research, University of Arizona.

Ill

References

Angell, J.K. and J. Korshover. 1985. Surface temperature changes following the six major
volcanic events between 1780 and 1980. Journal of Climate and Applied Meteorology
24:937-951.

Baldwin, B.B., J.B. Pollack, A. Summers, O.B. Toon, C. Sagan and W. van Camp. 1976.
Stratospheric aerosols and climatic change. Nature 263:551-555.

Deirmendjian, D. 1973. On volcanic and other particulate turbidity anomalies. Advances in
Geophysics 16:267-296.

Ellsaesser, H.W. 1986. Comments on "Surface temperature changes following the six major
volcanic episodes between 1780 and 1980". Journal of Climate and Applied Meteorology
25:1184-1185.

Fritts, H.C. 1974. Relationships of ring widths in arid-site conifers to variations in monthly
temperature and precipitation. Ecological Monographs 44:41 1-440.

Fritts, H.C. 1976. Tree Rings and Climate. Academic Press, London. 567 pp.

Fritts, H.C. (in press). Reconstructing Large-Scale Climatic Patterns from Tree-Ring Data: A
Diagnostic Study. University of Arizona Press.

Fritts, H.C. and D.J. Shatz. 1974. Selecting and characterizing tree-ring chronologies for
dendroclimatic analysis. Tree-Ring Bulletin 35:31-40.

Fritts, H.C. and J.M. Lough. 1985. An estimate of average annual temperature variations for
North America, 1602-1961. Climatic Change 7:203-224.

Fritts, H.C, G.R. Lofgren and G.A. Gordon. 1979. Variations in climate since 1602 as
reconstructed from tree rings. Quaternary Research 12:18-46.

Gilliland, R.L. 1982. Solar, volcanic and C02 forcing of recent climatic changes. Climatic
Change 4:111-131.

Gilliland, R.L. and S.H. Schneider. 1984. Volcanic, CO: and solar forcing of northern and
southern hemisphere surface temperatures. Nature 310:38-41.

Gordon, G.A. 1982. Verification of dendroclimatic reconstructions. In: Climate from Tree
Rings, M.K. Hughes, P.M. Kelly, J.R. Pilcher and V.C. LaMarche (eds.). Cambridge
University Press, Cambridge, pp. 58-61.

Gordon, G.A., J.L. Lough, H.C. Fritts and P.M. Kelly. 1985. Comparison of sea-level
pressure reconstructions from western North American tree rings with a proxy record of
winter severity in Japan. Journal of Climate and Applied Meteorology 24:1219-1224.

Hammer, C.U., H.B. Clausen and W. Dansgaard. 1980. Greenland ice sheet evidence of
post-glacial volcanism and its climatic impact. Nature 228:230-235.

112

Hirschboeck, K.K. 1979-80. A new worldwide chronology of volcanic eruptions.
Palaeogeography , Palaeoclimatology and Palaeoecology 29:223-241.

Hughes, M.K., P.M. Kelly, J.R. Pilcher and V.C. LaMarche. (Editors). 1982. Climate from
Tree Rings. Cambridge University Press, Cambridge. 223 pp.

Hunt, B.G. 1977. A simulation of the possible consequences of a volcanic eruption on the
general circulation of the atmosphere. Monthly Weather Review 105:247-260.

Kelly, P.M. and C.B. Sear. 1984. Climatic impact of explosive volcanic eruptions. Nature
311:740-743.

Kutzbach, J.E. and P.J. Guetter. 1980. On the design of paleoenvironmental networks for
estimating large-scale patterns of climate. Quaternary Research 14:169-187.

Lamb, H.H. 1970. Volcanic dust in the atmosphere: with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society of London
266:425-433.

Landsberg, H.E. and J.M. Albert. 1974. The summer of 1816 and volcanism. Weatherwise
27:63-66.

Legrand, M. and R.J. Delmas. 1987. A 220-year continuous record of volcanic H2S04 in the
Antarctic ice sheet. Nature 327:671-676.

Lough, J.M. and H.C. Fritts. 1985. The Southern Oscillation and tree rings: 1600-1961. Journal
of Climate and Applied Meteorology 24:952-966.

. 1987. An assessment of the possible effects of volcanic eruptions on North American

climate using tree-ring data, 1602 to 1900 A.D. Climatic Change 10:219-239.

Lough, J.M., H.C. Fritts and Wu Xiangding. 1987. Relationships between the climate of
China and North America over the past four centuries: a comparison of proxy records. In:
The Climate of China and Global Climate: Proceedings of the Beijing International
Symposium on Climate, 30 October-3 November 1984, Beijing, China. Ye Duzheng, Fu
Congbin, Chao Jiping and M. Yoshino (eds.). China Ocean Press, Springer-Verlag
(foreign distributor). New York and Berlin, pp. 89-105.

Milham, W.I. 1924. The year 1816 - the causes of abnormalities. Monthly Weather Review
52:563-570.

National Academy of Sciences. 1975. Understanding Climatic Change. National Academy
Press, Washington, D.C. 239 pp.

National Science Foundation. 1987. Climate Dynamics Workshop on Paleoclimate Data-Model
Interaction. Climate Dynamics Program. National Science Foundation, Washington, D.C.

113

Newhall, C.G. and S. Self. 1982. The volcanic explosivity index (VEI): an estimate of explosive
magnitude for historical volcanism. Journal of Geophysical Research 87C2: 1231-1238.

Oliver, R.C. 1976. On the response of hemispheric mean temperature to stratospheric dust: an
empirical approach. Journal of Applied Meteorology 15:933-950.

Parker, D.E. 1985. Climatic impact of explosive volcanic eruptions. Meteorological Magazine
114:149-161.

Pollack, J.B., O.B. Toon, C. Sagan, A. Summers, B.B. Baldwin and W. van Camp. 1976.
Volcanic explosions and climatic change: a theoretical assessment. Journal of Geophysical
Research 81:1071-1083.

Rampino, M.R. and S. Self. 1982. Historic eruptions of Tambora (1815), Krakatau (1883), and
Agung (1963), their stratospheric aerosols, and climatic impact. Quaternary Research
18:127-143.

. 1984. Sulphur-rich volcanic eruptions and stratospheric aerosols. Nature 310:677-679.

Robock, A. 1981. A latitudinally dependent volcanic dust veil index and its effect on climate
simulations. Journal of Volcanology and Geothermal Research 11:67-80.

Sear, C.B., P.M. Kelly, P.D. Jones and CM. Goodess. 1987. Global surface temperature
responses to major volcanic eruptions. Nature 330:365-367.

Self, S., M.R. Rampino and J.J. Barbera. 1981. The possible effects of large nineteenth and
twentieth century volcanic eruptions on zonal and hemispheric surface temperatures.
Journal of Volcanology and Geothermal Research 1 1:41-60.

Stockton, C.W., W.R. Boggess and D.M. Meko. 1985. Climate and tree rings. In: Paleoclimate
Analysis and Modeling. A.D. Hecht (ed.). John Wiley and Sons, New York. pp. 71-151.

Stommel, H. and E. Stommel. 1983. Volcano Weather. The Story of 1816, the Year Without a
Summer. Seven Seas Press, Newport, Rhodes Island. 177 pp.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science 224:1191-
1198.

114

Volcanic Effects on Colorado Plateau Douglas-Fir Tree Rings

Malcolm K. Cleaveland1
Abstract

The explosion of Tambora in April 1815, is the largest volcanic eruption in recorded history.
Based on measured temperatures, the significant North American climatic effects of Tambora
appear to have been limited to the northeastern United States and eastern Canada. However,
diameter growth of conifers on the Colorado Plateau in the southwestern United States was
extremely large from 1815 to 1817, and the largest regionally-averaged late season (latewood)
growth in 491 years occurred in 1816. This abnormal growth is probably not coincidental.
Above-normal growth occurs when moisture stress is reduced, and the eruption probably resulted
in abnormally low growing-season temperatures and/or abundant precipitation over the Colorado
Plateau. Latewood density was above average from 1815 to 1817, also indicating reduced
moisture stress. Trees located on marginal sites that are usually subject to the greatest moisture
stress showed the most favourable growth response from 1815 to 1817. These growth changes
are postulated to be effects of abnormally cool growing seasons that reduced evapotranspiration
and delayed onset of drought-induced late summer dormancy. Delayed dormancy would favour
development of an anomalously large latewood zone and increased latewood density. No
comparable growth responses are apparent for other known large eruptions, indicating that
regional climatic response to volcanic forcing is highly variable. Long-lived, climatically-
responsive trees are widely distributed in the northern hemisphere. Analyses of these tree-ring
data during recent centuries when instrumental climatic data are sparse may help reveal the
impact of known volcanic eruptions on northern hemisphere climate, and may also help identify
and date extremely large prehistoric eruptions.

Introduction

The April 1815 explosion of Mount Tambora at Latitude 8°S was the "largest and deadliest
volcanic eruption in recorded history..." (Stothers 1984). Lamb's (1970, Tables 7a,b) Dust Veil
Index is larger for 1815 than any year since 1500. The 1815 eruption also has the highest
Volcanic Explosivity Index since 1500 (Newhall and Self 1982). If such volcanic eruptions do
affect global climate, then this huge eruption should have left evidence in instrumental climatic
records and perhaps in other proxy climatic records such as ice cores and tree rings (Shutts and
Green 1978; Bryson and Goodman 1980; Gilliland and Schneider 1984; Kelly and Sear 1984;
Angell and Korshover 1985; Bradley 1988).

Volcanic aerosols reduce solar radiation at the surface on a regional or global basis by increasing
the albedo of the upper atmosphere. Temperature effects of aerosols are more closely related to
the quantity of sulphates injected into the upper atmosphere than to the quantity of fine particulate
ejecta. Sulphuric acid created as a result of sulphur-rich volcanic eruptions plays a major role in
reducing transmission of direct solar radiation (Harshvardhan and Cess 1976; Pollack et al.
1976). If the Tambora eruption had a global climatic impact, it must have created a sulphuric acid
aerosol. In fact, the 1815 Tambora episode coincides with a very large acidity peak from 1815

Department of Geography, University of Arkansas, Fayetteville, Arkansas 72701, U.S.A.

115

to 1818 in Greenland ice cores, even before correction for losses in transport from equatorial to
polar latitudes (Hammer et al. 1980; Rampino and Self 1984; Stothers 1984).

Despite abundant evidence pointing to the 1815 Tambora eruption as having all the requisite
characteristics for a major influence on global climate, Stothers (1984) and Angell and Korshover
(1985) found little evidence of significantly lower surface temperatures following this eruption.
Most of the available long instrumental-temperature records that form the basis of this conclusion,
however, are confined to Europe and the northeastern United States. In addition, available
records show that a period of well-below-average temperature started at least five years before
1815, which may mask some of the volcanic effects (Stothers 1984; Angell and Korshover 1985;
Baron, this volume). While 1816 is cooler than 1815 in most of these records, the drop is small
compared to the general cooling shown in the period. The long temperature record from New
Haven, Connecticut does show signs of considerable cooling, in keeping with the New England
reputation of 1816 as "the year without a summer" (Stommel and Stommel 1983; Angell and
Korshover 1985), but the northeastern United States and eastern Canada (Wilson 1985) are
considered exceptions. Horstmeyer (1989) compiled Cincinnati, Ohio daily weather records from
1814 to the present and found that, "No year since [1816] has even come close to having such
a cold summer". This demonstrates for the first time that an abnormally cold summer of 1816
occurred in the American Midwest, as well as in eastern North America.

There is a large and growing network of old, climate-sensitive tree-ring chronologies available
that might provide more spatially complete evidence for volcanic effects on climate (Stockton et
al. 1985). One such network from western North America has been used to estimate seasonal and
annual temperature variation for the United States (Fritts and Lough 1985; Lough, this volume).
The reconstructed temperature estimates were then used to study average climatic response to
selected volcanic events (1602 to 1900) by comparing temperature before and after the events
with superimposed epoch analysis (Lough and Fritts 1987). After low-latitude eruptions like
Tambora, cooling was especially pronounced in the spring and summer during the trees' growing
season. The spatial effects on United States climate were often different, and directly out of
phase, between the west coast and the rest of the country. During the spring following eruptions,
the Pacific Northwest experienced warming while the rest of the country cooled down. In summer
the area of warming expanded down the west coast, while the central and eastern United States
remained relatively cool. Lough and Fritts (1987), however, did not specifically report on the
Tambora eruption. In this paper a set of climate-sensitive tree-ring width and density chronologies
from the Colorado Plateau are used to investigate the possible impact of the 1815 eruption on the
climate of the southwestern United States.

Methods

Three sets of Douglas-fir (Pseudotsuga menziesii) tree-ring radial samples were collected and
crossdated with standard methods (Stokes and Smiley 1968). The total number of radii in the
collections ranged from 14 to 20, taken from 10 to 14 trees per site. The samples were collected
at Ditch Canyon (DIT) on the Colorado/New Mexico border, at Mesa Verde National Park in
Spruce Canyon (SPC) during 1978 (Cleaveland 1983, 1986, 1988) and in Bobcat Canyon (BOB)
in 1972 (Drew 1976) (Figure 1).

116

109° 108° 107°

J ! I

38° -

37°-

N
I

SPC

BOB* DIT^ Colorado

~ New Mexico

0 20

0 30

Figure 1: Map of the three sites sampled: Bobcat Canyon (BOB), Spruce Canyon (SPC), and Ditch
Canyon (DIT).

Conifer tree rings are divided into earlywood and latewood zones, formed first and last in the
growing season, respectively. Earlywood formation is most strongly influenced by climate in the
spring and latewood formation by late spring to mid-summer climatic conditions. Typical
earlywood cells become large, with relatively thin cell walls surrounding large cavities or lumens.
Typical latewood cells are smaller than earlywood cells, with thick cell walls and small lumens.
For this reason the earlywood part of a ring is less dense than the latewood portion. The
transition to latewood is abrupt in Douglas-fir (Panshin and de Zeeuw 1970). The width of the
two zones was measured optically from the BOB and DIT specimens. Characteristics of the SPC
samples, including latewood width and average latewood density, were measured by X-ray
densitometry (Parker et al. 1980; Cleaveland 1983, 1986).

Time series of ring widths are often not statistically stationary because the mean and variance
may both change with increasing age and diameter of the tree. The most common form of the
growth function approximates an exponential curve declining to a constant value, but linear
regression lines or more flexible polynomial or spline curves are also often used to remove

117

growth trend (Stokes and Smiley 1968; Fritts 1976; Cook and Peters 1981). To transform the
measurements into stationary time series, a curve is fitted to the measurements from each sample,
and each annual value is divided by the corresponding annual curve value. This transforms
measurement series into indices with a mean of 1.0, removing the effects of differences in mean
growth from tree to tree, and rendering the variance quasi-stationary. The indices for each radial
series from a site are averaged on an annual basis into a site chronology. The site chronology has
a mean equal to 1 .0 and a minimum value greater than 0.0, and represents a selected statistical
sample of the macro-environmental factors that control the radial growth of a given species on
a certain site through time.

Results and Discussion

BOB and DIT are lower forest-border sites that often experience high levels of moisture stress,
whereas the SPC site is more mesic (Drew 1976; Cleaveland 1983, 1986). One measure of
response to climate is the mean sensitivity statistic, that is, the average first difference of
chronology indices (Fritts 1976). The mean sensitivities of ring-width chronologies at BOB, DIT,
and SPC are 0.45, 0.44, and 0.28, respectively. This statistic indicates that the BOB and DIT
chronologies should show greater response to departures from normal growing-season conditions
than the SPC chronology.

A width or density index greater than 1.0 indicates above average growth that is usually
attributable to a cool and/or moist growing season in the Southwest (Fritts et al. 1965; Fritts
1976). When latewood width at the BOB, DIT, and SPC sites are averaged for each year, the
average index is larger for 1816 than for any year since 1487, a period of almost five centuries
(Figure 2). In addition, the ring-width, latewood-width, and latewood-density indices for the three
collection sites all equal, or greatly exceed, average growth (1.00) for 1815, 1816, and 1817
(Table 1). These anomalies indicate that the growing seasons were substantially cooler and/or
wetter than normal (Cleaveland 1983, 1986).

The very large values of latewood growth in the decade of the 1490s (Figure 2) are probably
artifacts of a small number of samples, and end effects of curve fitting. The best replicated
chronology, SPC, has an index of only 1.33 in 1491 (Figure 2). If the poorly-replicated 1490s
are ignored, 1815 and 1816 summed are the largest average latewood total of two consecutive
years, and 1816 and 1817 are the second largest. In addition, if the 1490s are not considered, the
1815-17 period has the largest total latewood growth of three consecutive years.

All chronologies are well replicated after about 1700, giving greater confidence in the estimated
index means after 1700. It would certainly be possible to increase the sample depth of long series
at many sites in the western United States to improve estimates in the early part of the
chronologies. This should be an important consideration before using these chronologies to
investigate earlier volcanic eruptions.

Lower forest-border Douglas-fir trees in southwestern Colorado generally become dormant in
June or July - forced into dormancy by moisture stress long before photoperiod or low
temperatures could become responsible (Fritts et al. 1965). Also, a conditional probability
analysis of 89 southwestern conifer chronologies (e.g., Stockton and Fritts 1971) indicates that
the influence of temperature on tree growth late in the growing season is stronger than the
influence of precipitation (Cleaveland, unpublished data). Chronologies at those sites showing the
highest degree of inferred moisture stress (BOB and DIT) show a stronger response to the

118

1815-17 climatic anomaly than the more mesic SPC site. It seems probable, therefore, that the
growth anomaly is linked in some way to below-normal temperature and/or above normal
precipitation that drastically reduced moisture stress on Colorado Plateau trees during those
growing seasons. The greatly enlarged latewood zone in the 1815-17 rings of these Colorado
Plateau conifers could be interpreted as evidence for a longer-than-normal growing season
extended by below-normal air temperatures during the summer. Normal or above-average
precipitation probably also occurred during the extended growing seasons from 1815 to 1817.

Table 1: Southwestern Colorado Tree-Ring Chronology Indices (1810-20).'

Bobcat Canyon

Ditch Canyon

Spruce Canyon

Average

Ring

Latewood

Latewood2

Latewood

Year

Width

Density

Width

1810

0.68

0.70

0.77

1.03

1.01

0.83

1811

1.01

1.07

2.14

1.32

1.05

1.51

1812

1.12

0.87

1.09

0.87

1.01

0.94

1813

0.41

0.36

0.75

0.31

0.88

0.47

1814

0.92

1.03

0.69

0.97

0.88

18153

1.24

1.73

2.29

1.50

1.06

1.84

1816

2.12

3.47

4.58

1.57

1.10

3.20

1817

2.28

2.23

1.78

1.00

1.02

1.67

1818

0.47

0.42

0.28

0.83

0.97

0.51

1819

0.31

0.77

0.65

0.68

0.90

0.70

1820

0.24

0.32

0.64

0.34

0.86

0.43

' Indices greater than 1.0 indicate above-average growth, and indices less than 1.0 represent
below-average growth.

2 Latewood density variability was multiplied by 3.0 to increase the range of variation relative
to the other variables.

3 The year Tambora erupted. Table 1

Other historic eruptions are believed to have affected climate, and might have influenced the
growth of trees on the Colorado Plateau. Rampino and Self (1984) list selected eruptions with
estimates of the sulphuric acid aerosol generated and the estimated northern hemispheric
temperature change. The eruption of Laki in 1783 is estimated to have caused greater cooling
than Tambora, but no effect can be detected in Figure 2. Laki is a high-latitude (64°N) volcano,
however, and Lough and Fritts (1987) found that volcanic eruptions in low latitudes resulted in
the greatest climatic response across the United States.

There appears to have been no growth response of Colorado Douglas-fir to Krakatau (6°S),
unless there was a weak effect delayed until 1885. The eruption of Santa Maria (15 °N) occurred
in 1902, a year of intense drought on the Colorado Plateau (Cleaveland 1983, Appendix 1). The
growing season of 1903 had adequate precipitation, and growth was slightly above average, but
1904 was very dry resulting in low growth (Figure 2). The lack of adequate precipitation would

119

certainly curtail the possible response of tree growth to volcanic cooling. Climatic effects from
the eruptions of Katmai (58°N) in 1912 and Agung (8°S) in 1963 are also not discernible in these
chronologies (Figure 2). The growth anomaly at 1816 is clearly the largest apparent in these data,
and is probably the only one that can definitely be attributed to a known major volcanic eruption.
However, there are other pronounced increases in growth that may be associated with
undocumented volcanic activity. The possible detection of other volcanically created climatic
effects in Colorado Plateau tree growth deserves further study.

4 -
3 -

2 -
1

2
1
5
4
3
2

1 -
4

3 -

2 -
1 -
2

1 H
o

n 1 1 1 1 1 1 1 1 r

1000 1650

1 I 1 1 — ' — 1 I 1
1700 1750

YEAR

-i — i — i — r~

1487 ISOO

1850

1950 1978

Figure 2: Plots of tree-ring chronology index series from southwestern Colorado. A = average latewood
width from Bobcat, Ditch, and Spruce Canyon sites; B = Ditch Canyon latewood width; C =
Spruce Canyon latewood width; D = Bobcat Canyon latewood width; E = Bobcat Canyon ring
width; F = Spruce Canyon latewood density (with variability multiplied by 3.0 to increase
variability relative to the other variables). The numbers above each chronology are the sample
depth at that point.

The atmospheric mechanisms that may create regional climatic anomalies in response to volcanic
influences are not well understood. Atmospheric and oceanic conditions at the time an eruption
occurs may determine climatic response. It is believed, for example, that sea-surface temperatures
and the El Nino-Southern Oscillation phenomenon have mediated climatic effects of several recent
volcanic eruptions (Angell and Korshover 1985; Angell 1988).

Conclusions

The use of moisture-stressed conifers to investigate spatial patterns of historic volcanic eruption
effects on climate may partially compensate for the limited distribution of instrumental climatic
records prior to the twentieth century. Latewood width is particularly sensitive to the growing
season moisture budget. The 1816 annual rings investigated in this report have the largest amount
of latewood growth on the Colorado Plateau in 491 years. The pattern of greatly increased ring
width, latewood width, and latewood density in these Colorado Plateau conifers from 1815 to
1817 indicates a reduction of growing season moisture stress unique in the last five centuries. The

120

climatic effect that apparently reached a maximum over the Colorado Plateau in 1816 probably
began shortly after the April 1815 eruption of Tambora, and persisted into the growing season
of 1817. The cause of this extraordinary growth anomaly was probably a reduction of mid-
summer evapotranspiration demand, which appears to have extended the growing season,
resulting in extremely large latewood growth. No effects of other known large volcanic eruptions
were detected in these tree-ring chronologies. The receptivity of the general circulation to
volcanic forcing may partly explain the apparently strong climate-tree growth response on the
Colorado Plateau to the Tambora eruption, and the absence of a large growth response to other
major eruptions during the historic period. Effects of eruptions on regional temperature and tree
growth might be masked by existing regional climatic conditions such as drought, or by other
climatic-forcing mechanisms such as sea-surface temperatures and/or the phase of the El Nino-
Southern Oscillation. This study has focused on annual ring data from a small set of chronologies
in a small part of the Colorado Plateau, but it demonstrates a potential application of tree-ring
data to the analysis of volcanic effects on climate.

Acknowledgements

Thanks are due David W. Stahle, University of Arkansas Tree-Ring Laboratory, for suggested
improvements to the manuscript, and to Thomas Harlan, University of Arizona Laboratory of
Tree-Ring Research, who assisted in the collection and dating of the samples. Part of the data
comes from my doctoral dissertation, which was supported by the Laboratory of Tree-Ring
Research and the Department of Geosciences, University of Arizona. Additional support from
the National Science Foundation, Climate Dynamics Program (grant ATM-8612343) is also
acknowledged.

References

Angell, J.K. 1988. Impact of El Nino on the delineation of tropospheric cooling due to volcanic
eruptions. Journal of Geophysical Research 93D:3697-3704.

Angell, J.K. and J. Korshover. 1985. Surface temperature changes following the six major
volcanic episodes between 1780 and 1980. Journal of Climate and Applied Meteorology
24:937-951.

Bradley, R.S. 1988. The explosive volcanic eruption signal in northern hemisphere continental
temperature records. Climatic Change 12:221-243.

Bryson, R.A. and B.H. Goodman. 1980. Volcanic activity and climatic changes. Science
207:1041-1044.

Cleaveland, M.K. 1983. X-ray densitometric measurement of climatic influence on the intra-
annual characteristics of southwestern semiarid site conifer tree rings. Ph.D. dissertation.
University of Arizona, Tucson. 177 pp.

. 1986. Climatic response of densitometric properties in semiarid site tree rings. Tree-Ring

Bulletin 46:13-29.

121

. 1988. Corrigendum to: Climatic response of densitometric properties in semiarid site

conifer tree rings. Tree-Ring Bulletin 48:41-47.

Cook, E.R. and K. Peters. 1981. The smoothing spline: a new approach to standardizing forest
interior tree-ring width series for dendroclimatic studies. Tree-Ring Bulletin 41:45-53.

Drew, L.G. (ed.). 1976. Tree-ring Chronologies for Dendroclimatic Analysis. University of
Arizona Laboratory of Tree-Ring Research, Tucson. 64 pp.

Fritts, H.C. 1976. Tree Rings and Climate. Academic Press, London. 567 pp.

Fritts, H.C, D.G. Smith and M.A. Stokes. 1965. The biological model for paleoclimatic
interpretation of Mesa Verde tree-ring series. American Antiquity 31:101-121.

Fritts, H.C. and J.M. Lough. 1985. An estimate of average annual temperature variations for
North America, 1602 to 1961. Climatic Change 7:203-224.

Gilliland, R.L. and S.H. Schneider. 1984. Volcanic, C02 and solar forcing of northern and
southern hemisphere surface air temperatures. Nature 310:38-41.

Hammer, C.U., H.B. Clausen and W. Dansgaard. 1980. Greenland Ice Sheet evidence of post-
glacial volcanism and its climatic impact. Nature 288:230-235.

Harshvardhan and R.D. Cess. 1976. Stratospheric aerosols: effect upon atmospheric temperature
and global climate. Tellus 28:1-9.

Horstmeyer, S.L. 1989. In search of Cincinatti's weather. Weatherwise 42:320-327.

Kelly, P.M. and C.B. Sear. 1984. Climatic impact of explosive volcanic eruptions. Nature
311:740-743.

Lamb, H.H. 1970. Volcanic dust in the atmosphere; with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society of London
A266: 425-533.

Lough, J.M. and H.C. Fritts. 1987. An assessment of the possible effects of volcanic eruptions
on North American climate using tree-ring data, 1602 to 1900 A.D. Climatic Change
10:219-239.

Newhall, C.G. and S. Self. 1982. The volcanic explosivity index (VEI): an estimate of explosive
magnitude for historical volcanism. Journal of Geophysical Research 87C: 1231-1238.

Panshin, A.J. and C. de Zeeuw. 1970. Textbook of Wood Technology, Vol. I, Third Edition.
McGraw-Hill Inc., New York. 705 pp.

Parker, M.L., R.D. Bruce and L.A. Jozsa. 1980. X-ray densitometry of wood at the W.F.P.L.
Forintek Canada Corporation Western Laboratory, Technical Report No. 70:1-18.

122

Pollack, J.B., O.B. Toon, C. Sagan, A. Summers, B. Baldwin and W. Van Camp. 1976.
Volcanic explosions and climatic change: a theoretical assessment. Journal of Geophysical
Research 81:1071-1083.

Rampino, M.R. and S. Self. 1984. Sulphur-rich volcanic eruptions and stratospheric aerosols.
Nature 310:677-679.

Shutts, G.J. and J.S.A. Green. 1978. Mechanisms and models of climatic change. Nature
276:339-342.

Stockton, C.W. and H.C. Fritts. 1971. Conditional probability of occurrence for variations in
climate based on width of annual tree rings in Arizona. Tree-Ring Bulletin 31:3-24.

Stockton, C.W., W.R. Boggess and D.M. Meko. 1985. Chapter 3, Climate and tree rings. In:
Paleoclimate Analysis and Modelling. A.D. Hecht (ed.). John Wiley and Sons, New
York. pp. 71-161.

Stokes, M.A. and T.L. Smiley. 1968. An Introduction to Tree-Ring Dating. University of
Chicago Press, Chicago. 73 pp.

Stommel, H. and E. Stommel. 1983. Volcano Weather. Seven Seas Press, Newport, Rhode
Island. 177 pp.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science 224:1191-
1198.

Wilson, C. 1985. The Little Ice Age on eastern Hudson/James Bay: the summer weather and
climate at Great Whale, Fort George and Eastmain, 1814-1821, as derived from Hudson's
Bay Company records. In: Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus
55:191-218.

123

1816 in Perspective: the View from the Northeastern United States

William R. Baron1
Abstract

The year 1816 is remembered in the northeastern United States as one of the harshest, coldest
years on record, and continues even to the present to be one of the most widely known folk-
climate episodes of the region. A study of the period 1790-1839 helps to place 1816 in its
climatological context. New evidence supports the case that 1816 had a particularly cold and dry
growing season but was by no means the coldest or driest year of the period. Several other years
during the second and fourth decades of the century climatically rivaled the abnormal conditions
of the "year with no summer". 1816's claim to fame rests on the severe impact that the cold and
drought had on the area's then extensive agricultural operations. For this reason 1816 came to
represent the abnormal climatic conditions of not only a single year but also most of the second
decade of the nineteenth century.

Introduction

The year with no summer, 1816, is one of the best known folk-weather occurrences of the
northeastern United States. During the nineteenth century, "eighteen-hundred-and-froze-to-death"
was a subject for many newspaper articles, autobiographical reminiscences, and local histories
(Mussey and Vigilante 1948). Even now, some 170 years after that frosty summer, it continues
to be a topic of great popular interest, still commanding feature articles in the region's
newspapers and periodicals (Fichter 1971; Leach 1974; Reichmann 1978; Parsons 1980).

1816 also has not escaped the attention of scientific investigators, and has been a subject of
considerable debate since the early nineteenth century (Skeen 1981). During the twentieth
century, research appearing on this topic included that by W.I. Milham (1924), J.B. Hoyt (1958),
H.E. Landsberg and J.M. Albert (1974), H. and E. Stommel (1979, 1983), and R.B. Stothers
(1984), and centred on the issue of what factors contributed to 1816's abnormal summer. Most
researchers concluded that the great Tambora eruption of 1815 and low sunspot activity were the
major factors involved; although a minority, including Landsberg, have questioned the influence
of volcanic dust in the atmosphere as a major contributor. Of late, historians have begun to assess
the climatic impact of the 1810s on society in both Europe and North America (Post 1977; Skeen
1981). Our continuing interest in and fascination with 1816 finally led to the international
conference that produced the papers included in this volume.

The purpose of this study is to present new and additional evidence for the northeastern United
States covering the period 1790 through 1839, in order to help place 1816 in its proper
climatological context. Data presented include instrument readings for temperature, precipitation,
and wind direction. Additional reconstructions for snowfall, seasonal precipitation, cloud cover,
growing-season length, thunderstorm frequency, river freeze-up and ice-out, and phenology
records (based on the analysis of qualitative materials such as diaries), weather journals and
newspaper reports, also are discussed. However, before proceeding, I will present a

Historical Climate Records Office, Center for Colorado Plateau Studies, Northern Arizona University, Box 5613,
Flagstaff, Arizona 86011, U.S.A.

124

reconstruction of the weather history of 1816 based on the observations of 74 diarists from
northeastern United States.

1816, "The Year With No Summer-
Appropriately enough, 1816 began, at least in Phillipstown, Massachusetts, with enough snow
on the ground for sleighing. All over New England, January was a snowy, stormy month until
the very end when a sudden thaw caused localized flooding such as the one reported by Isaiah
Thomas at Worcester, Massachusetts on 23 January where some mill dams were carried off and
some items stored in a warehouse were destroyed.

According to among others, Leonard Hill of East Bridgewater, Massachusetts, February was a
mild and pleasant month with only three snows reported. By the beginning of March there was
little deep snow anywhere with the exception of most of northern New England. Early March was
clear and cold, and was followed by a series of three snow storms around mid-month that
produced a few days of sleighing but soon melted. On 28 and 30 March, warm air returned
producing thunder and lightning as reported by Elijah Kellogg at Portland, Maine and Thomas
at Worcester.

April quickly turned cold again with frequent frosts and some snow. However, by 14 April, there
was little snow left at Hallowell, Maine. By 19 April, Alexander Miller of Wallingford, Vermont
had begun to plough his fields; Stephen Longfellow of Gorham, Maine was already planting
wheat; and Theodore Lincoln of Dennysville (in far downeast Maine) was reporting ice-out on
the local streams - a sure sign of coming spring. At the end of the month, Joshua Lane of
Sanbornton, New Hampshire already was reporting the start of a drought that would later plague
all of northern New England.

In early May, farmers throughout the region completed planting their major crop, corn (maize).
By mid-month the weather had become "backward" again with a "heavy black frost" that froze
the ground to at least one-half inch (1.3 cm) reported on 15 May as far south as Trenton, New
Jersey. Miller, at Wallingford, Vermont, reported snow on 14, 17 and 29 May while Lane, over
at Sanbornton, saw a large frost on 24 May, and ended the month with further complaints about
the continuing drought. B.F. Robbins, visiting Concord, New Hampshire noted that May ended
with two days of "remarkable cold" that froze the ground "to near an inch".

June is the month most remembered for its outbreak of cold weather. On 4 June, there were
frosts at Wallingford, Vermont and Norfolk, Connecticut. By 5 June the cold front was reported
over most of northern New England. On 6 June, snow was reported at Albany, New York and
Dennysville, Maine, and there were killing frosts at Fairfield, Connecticut.
7 June brought reports of severe killing frosts from across the region, and as far south as
Trenton, New Jersey.

Typical of comments by diarists concerning this day are those by George W. Featherstonhaugh
of Albany, New York, who wrote that the frost killed most of the fruit, as many apple trees were
then just finishing blossoming. Leaves on most of the trees were "blasted" by the cold. Corn and
vegetable crops were injured. He also feared many of the sheep that had just been sheared might
die of cold.

Cold weather continued through the night of 10 June. By the end of the month most observers
were reporting the return of warm weather, but by then most crops were either killed or

125

"backward" and stunted in their growth. In northern New England, those crops that survived the
frosts were hit by what was now a very serious drought, greatly reducing production of one of
the area's primary crops, hay.

In early July there was another outbreak of cold weather in northern New England. On 5 July,
at Gorham, Maine, there was a very hard frost. Benjamin Kimball of Concord, New Hampshire
and Thomas Robbins of Norfolk, Connecticut reported hard frost on 7 July. There was frost on
8 July at Portland, Maine and on the following day at Sanbornton, New Hampshire. Thereafter
the cold held off for the remainder of the month. Throughout the entire month dry conditions,
generally reported earlier in northern areas, persisted.

Frosts returned on the morning of 21 August, being reported at York and Portland, Maine and
Wallingford, Vermont. By 22 August hard frosts were noted all over the region and as far south
as Trenton where buckwheat crops were killed. Thomas, at Worcester, Massachusetts, reported
that these frosts "cut off Indian corn in many places", while others such as Hill at East
Bridgewater, Massachusetts observed that frosts did little or no damage.

The frosts continued into September. In northern New England there were frosts on 10 and 11
September and throughout New England during 25-27 September. On 28 September, there was
a killing frost throughout the region extending as far south as Trenton. It killed any vegetation
that had somehow survived to that date. The drought in northern New England was finally broken
by rains in the last week of the month.

The remainder of autumn was very mild with very few snowfalls or storms. December was also
mild, until the last 10 days or so, when it turned cold enough to freeze the harbour at Beverly,
Massachusetts. The year ended with enough snow on the ground at Phillipstown, Massachusetts
to use a sleigh.

The place of 1816 in the memory of the regional population has been summed up well by the
historian H.F. Wilson when he wrote that, in 1816, farmers experienced an "almost total failure"
of major crops. There was a fair yield of winter grain, but other crops such as corn and hay
failed leading to the loss of many sheep and cattle for lack of feed during the following winter.
As a result 1816 has come down to us as the "cold year", "the famine year" and "eighteen
hundred and froze to death".

Of great interest to climatologists and historians alike is the fact that 1816 was not the only
difficult, abnormal year of the second decade of the nineteenth century. Based on statistical
analysis of climatological data, other years might justifiably claim a portion of 1816's notoriety.
An analysis of the 50 years surrounding 1816 serves to locate some of these years and to place
the "year with no summer" in its climatological context.

Databases and Methodologies

The analysis that follows is based upon records assembled from several databases. The first of
these is a set of eight yearly mean temperature records comprised of instrument readings for
periods of 26 years or more, that overlap 1816 by at least three years and that, in composite,
cover the 50 years from 1790 through 1839. From north to south these records include: Castine,
Maine, 1809-39 (Baron et al. 1980); Brunswick, Maine, 1807-36 (Cleaveland 1867); Salem,
Massachusetts, 1790-1829 (Holyoke 1833); New Bedford, Massachusetts, 1813-39 (Rodman

126

1905); Williamstown, Massachusetts, 1811-38 (Milham 1950); New Haven, Connecticut, 1790-
1839 (Landsberg 1949); New Brunswick, New Jersey, 1790-1839 (Reiss et al. 1980); and
Philadelphia, Pennsylvania, 1790-1839 (Landsberg et al. 1968). All but one of these stations,
Williamstown, are located in the present coastal climatic zone as computed by the United States
National Oceanic and Atmospheric Administration. Williamstown's inland situation makes it more
vulnerable to outbreaks of Arctic cold coming from the northern interior of the continent.

Instrumental records of precipitation for the period are more limited than those for temperature.
The three best include those for New Bedford, 1814-39 (Rodman 1905); New Haven, 1804-21
(Kirk 1939); and Philadelphia, 1790-1839 (Landsberg etal. 1968). No long precipitation records
for inland locations are available.

To assure the representativeness of these long-run temperature and precipitation records, and to
enhance the record density and distribution throughout the region, a database of short-duration
instrumental records was assembled. The location of these records is shown in Figure 1. The
number of records available increases, by decade, in a steady progression from the 19 available
in the 1790s to 57 for the 1830s.

Yet another database (qualitative materials from diaries, journals, newspapers, and local histories
- many available only in manuscript form) was compiled to supplement the instrumental records.
At that time, only qualitative materials for New England were sufficiently organized for
inclusion. The database is comprised of 174 sources representing 55 of New England's 67
counties. Coastal and intermediate interior locations are well represented whereas data for some
upland and western interior counties are missing. From this database, frequency counts for days
with precipitation, fair skies, thunder and lightning storms, westerly winds, and snowfalls, as well
as the yearly dates for spring and autumn killing frosts, length of snowfall seasons, dates of
apple-tree blossoming and lengths of droughts were compiled. The methodology used to
reconstruct these various records is explained in Baron (1988, 1989), Baron and Gordon (1985)
and Baron et al. (1984).

Compilations of killing-frost reports were further refined by computing the year and day of each
frost and calculating the length of the growing season. To assess the possible impact of these
frosts on a major crop, information from an agricultural database made up of over 60 farm
journals was used to provide the mean dates for the planting and harvesting of Indian corn.

Analysis of Records

As can be seen in part from Figure 2, moving from north to south, 1816's yearly mean
temperature is not the coldest for the period. In Maine it was only the fourth lowest; while farther
south at Salem, New Haven, and New Bedford it was third or second lowest for the record. Even
farther south at New Brunswick and Philadelphia, 1816 was close to the mean for the period 1790
through 1839. Farther inland at Williamstown, 1816 was the seventh coldest year (Figure 3). In
New England the years noted to be as cold or colder than 1816 include: 1790, 1812, 1817, 1818,
1823, 1835, 1836 and 1837. 1836 and 1812 appeared most often as the two coldest years. South
of New England, 1836 and 1817 were the coldest years.

127

Houllon «\

Dennysville ©

III

Dgdensburg
©Gouverneur

l^jBurlingtoi

LVermonl

Rutland

> Hanover

Derryfield
o o

I Concord g

. CCASTINE
Gardiner

1 "V □

« " 1

°BRUNSWICK

oHiram f

e Portland
o^Saco

Dover

Pultneyville

Utica*

Manlius0

Rochester

'Portsmouth
. P'n9

New Hampshire

k ^^Watertown*

iie-mu^ V Cambridge-TCharlestown

WILLIAMSTOWN^--prince,on ydBjBoston
to ° ^Dedham

Albany* Amherst « Medtieldo Canton lMlaSSaChUS©ttS
Mendon ^".Leicester^-j

lenceJl_ ■***
\ JOi. ©NEW BEDFORD

-Provider

NewpoTfcV- jT> *Nan|ucke|

Carlislee

I Brooklyn
^elleville© filNew York
Chatham o JrGovemor's Is

Nazareth •

"NEW BRUNSWICK

_ lew Jersey

Germantown y^Burlington
Springfield 4 ^PHILADELPHIA

LMend
Ellington o
Hartford «

ConnecticuJjji.Rhode Island
Hfmden>NEw^Avr *

Chatham©

Lancaster
Gettysburg «

ure 1: Northeastern United States. Instrumental record locations, 1790-1840.

128

cr> ■»— colo r-. o> t— cocor— cr> »— co u"> r-- o> -r— co uo cr> ■»— co in r-- o>

cocn ct>o*>ct>o> ooo oo CNJCvJOJOJCVJ COCO CO CO CO xj-

■ . ■ - r-r- r-* go co co co co co go co oo co go co go go oo coco coco coco

Year

COO>010>0>0>00000»— '-CVJCMCNJCJCVJCOCOCOCOCOrr
r«»t^r-.i--r— r^cocococooococoaocooocooocooocooooooooococo

Year

o> co to r>- o> y— co i/> r-^ o> comr^-cn co in C7> co in r-- o> ■*—

CO O) 0> <7> CT> CT> O O OOO -w— v— ^— t~ C\J CM CM CM <\J CO CO CO CO CO

r-» r-- i**- r— r— r~-» co go co go oo co go co go go co go co oo co go go co co co co

D— Annual Mean Temperature Year — — 5 Year Running Mean

Figure 2: Castine, Maine (A); Salem, Massachusetts (B); and New Brunswick, New Jersey (C); annual
mean temperatures, 1790-1840.

129

Figure 3: Williamstown, Massachusetts. Yearly mean temperatures, 1811-38.

Talcing all temperature records together for the 50-year span, apparently the years 1790 through
1798 were either normal or slightly cooler than normal. 1799 through 1805 was warmer than
normal. A period of variability between 1806 and 1811 followed. Starting in 1812 (with some
variation until 1818) it was much cooler than normal. This cool period was followed by a series
of variable years from 1819 through 1823. From 1824 through 1830 it was somewhat warmer
than normal, and from 1831 through 1837 it was again very cool. The last two years of the
decade show a warming trend that extended into the 1840s.

Figure 4 shows an analysis of seasonal mean temperatures for Brunswick and New Haven. Winter
mean temperatures were calculated from December, January and February monthly means; spring
means from March, April and May; summer means from June, July and August; and autumn
means from September, October and November. The Brunswick record shows the winter of 1816
was very cold, whereas the summer was cool, the spring about average and the autumn a little
warmer than average. Summer mean temperatures in 1812 were as low as those for 1816. New
Haven presents a somewhat different picture with a mild winter, average autumn and cool spring
and summer. As far as New Haven is concerned, the summer mean temperatures for 1812 and
1817 were far lower than that for 1816.

Figure 5, a daily mean temperature record for 1816 kept at Brunswick, Maine shows why this
year is so well remembered. The outbreaks of cold in much of May, early June, early July, late
August and late September tell much of the story. These cold periods, all well below the monthly
mean temperature for the entire record (1807-36), doomed many a farmer's crops to failure. The
key to 1816's infamy lies in the extreme shortness of its growing season - the primary reason
why it, and not 1812 or 1836, has gone down in the regional weather lore as "the year with no
summer".

130

131

Daily Mean

Year Days

Monthly Means lor Record

0 30 60 90 120 150 180 210 240 270 300 330 360
' ■ ' ■ l i i i l i l i l i l i I i I i l i l_

Jan. Feb. March Apr May June July Aug. Sept. Oct. Nov. Dec.

Figure 5: Daily mean temperatures at Brunswick, Maine for 1816.

The plots of growing-season lengths for eastern Massachusetts, southern New Hampshire and
southern Maine (Figure 6) leave one unmistakable impression - 1816, by far, has the shortest
growing season. Other particularly short growing seasons occurred in 1808, 1824, 1829, 1834
and 1836. With the exception of 1816 and 1836, a number of these short seasons can probably
be attributed to one-night radiational cooling under clear skies during either spring or autumn,
and not to prolonged outbreaks of cold weather; otherwise these years would have appeared in
our lists of cool yearly and seasonal temperatures. Of course clear skies in combination with cold
fronts also contributed to some frosts during 1816, as that year has one of the highest percentages
of days with fair skies (Figure 7).

Figure 8, showing spring and autumn killing-frost dates in combination with corn-planting and
maturation dates, further illustrates the importance of growing-season data in understanding the
notoriety of 1816. For eastern Massachusetts, 1816 is the only year in which young corn was
killed in the spring after it had sprouted and in which corn that survived replanting was killed in
the autumn, before it could reach maturity. Under these circumstances, it is safe to assume that
in most places in New England corn crops were an almost total failure. The story for 1816 is the
same for New Hampshire and Maine. These reconstructions also show there were a number of
years when corn crops were hit by late spring or early autumn frosts. Particularly difficult periods
include: 1793-96, 1812-17, and 1835-36.

132

o~> i — co m r*- o> coior**»o> »— co ir> [--- o> co tr> r*- o> ■»— co to r-» en

C0<7>0> 0)0)05 0 OOOO ■»— f— CNiCNJCNJCNJ CNJ CO CO CO CO CO

r- r - r- r - r-~ r-- CO GO GO GO CO CO OO OO CO CO CO CO OO CO OO OO CO CO CO CO CO

Year

ct> co uor-«-o>"»— co U) n o> co un r-- o> •«— co in r-- o> co in r*- o> ■»—

CO CT> CJ> CJ> CT> CT> O OOOO i-r-i-i-t-W(\JCMW CvJ CO CO CO CO CO t}-

i — r — r-— I*-— r— r — co 00 co 00 00 co go go go go 00 00 go 00 00 go go go go go co

Year

cr> -» — co lo r-- o> co in r — cj> ■»— co in r-^. 01 co in r-» o> ^- co in r-- ct>

CO <J~> Cr> 0)00)00 O O O ■»— C\J CM CO CM CM CO CO CO coco^j-

r^— r— r-* r — r-» r — go go go 00 co co go go 00 co go 00 00 00 00 co go co co co co

■ — Length of Growing Season Year — — 5 Year Running Mean

Figure 6: Growing-season lengths for southern Maine (A), southern New Hampshire (B) and eastern
Massachusetts (C), 1790-1840.

133

100

aamroaoirooimaoooooooooO'-'-^'-^^^^^^fNiNNNNwwwwNoonnnonnonv
r^-r*>»rv~.r~--r~-rv~fv-rv-r — r^- (DcorococDtscocococococDCDCDcocDcDcDcDcocDcococooocococDcofficocooocDcDcococococD co

□ Winter

Spring

Year

Summer

Fall

Figure 7: Percentage of days with fair skies for eastern Massachusetts, 1790-1840.

Reconstructions indicating the onset of spring-like weather, such as dates of last snowfalls and
blossoming dates of various fruit trees such as apples (Figure 9), show that 1816's spring was
cooler and more unsettled than normal. However, it was far more satisfactory for agricultural
pursuits than 1812, 1832, or 1836 through 1838, when conditions were extremely "backward".

While the major part of the 1816 story lies in its growing-season record, there are several other
types of records in which 1816's position is worthy of mention. The first of these is the
precipitation record (Figure 10). 1816 was a year of about average precipitation, with the
exception of the summer, which was particularly droughty. Reconstructions concerning the period
and intensity of agriculturally-defined droughts show that the magnitude of dry conditions
increased significantly the farther north within the region one looks. For the 50 years from 1790
through 1840, the periods from 1791-1806 and 1813 to 1820 saw numerous growing-season
droughts. After 1820 the number of reported droughts decreased markedly. This apparent
increase in precipitation also can be seen in Figure 1 1 , illustrating the mean number of days per
year with precipitation for southern New England.

134

60 90 120 150 180 210 240 270 300 330

Year Day

60 90 120 150 180 210 240 270 300 330

Year Day

- Spring frost • — Corn Plant Date ■ — 120 Day Grow season * — Fall Frost

Figure 8: Killing frost and corn plant/harvest dates for southern Maine (A), southern New Hampshire
(B), and eastern Massachusetts (C), 1790-1840.

135

9 " ?

V II

i\ ii

ii ii

" II f

Year

B 160 ■

155 :

150 -j

145 -j

w 140 :
Q

« 135 :
>■

130 :

125 :
120 -j
115 :

110

il ii

<7>T— eo m r- o> ■»— w m n O) co in f-^. o> co m r*- o> co ir> o> i—

co cn cr> cn cn cn o OOOO *—■»—■»—■»— t— C\J CO OJ OJ Cvjcococo CO CO tJ-
r — r — r — r*>- go go go go go go go co go co go go go go go go go go co co go

Year

Figure 9: Apple-blossom dates for northern (A) and southern (B) New England, 1790-1840.

136

200

Year

Figure 10: New Bedford winter, spring, summer and autumn precipitation, 1814-40.

Year

oc\jrrtooooc\i-«j-tDoooc\j'^-tDa3ooj,«rtDoooc\jTrtoeoo

0>0><J>0>CT>OOOOOt— ■•— CMCMCVJCVJCMCOCOCDCOCO-"*-

r — r - r - (■•*■ r-- GO CO CO CO CO CO GO CO CO CO CO CO CO CO CO CO CO CO GO CO OO

160

140

120

~ 100

jjjjjjjjj^ y^p!

40 H

Figure 11: Mean number of days per year with precipitation for southern New England, 1790-1840.

137

r*» r — r-~ r*-. r-— r-. r — r — r-~ f"-- oo oo oo oo co co co oo co co co oo co co co go oo oo oo oo go oo go oo go oo oo go oo go go go go go oo go oo oo ao oo go

Year

= 30 -

Spring (Mar-May)
□ Winter (Dec-Feb)
■ Fall (Oct-Nov)

20 -

10 -

O T—

cn cn

c\jco^rinoDr^cocno^cvcr>Trin^^cocno^c\ico^incor^c»cno^c\jcoTfi^

oio>oioioioiO)OiooooooooooT-T-^^i-^^>-^->-(NiM<\wc\jw<MCMN(Njnnnrocoon
r — r-— r — r — r-- co co go oo co go co ao go go co go co go co oo co go go go go oo co co go go go go oo ao ao ao co ao co co co

MOOIO
CO CO CO ^J"
CO CO CO CO

Year

Figure 12: Number of days with snowfall in southern Maine (A) and eastern Massachusetts (B), 1790-
1840.

138

Snowfall records show that during the winter of 1815-16 there were relatively few days when it
snowed, especially in northern New England (Figure 12), but the New Bedford seasonal
precipitation record shows average or slightly above average precipitation for that winter.
However, under no circumstances can 1816 be viewed as a particularly snowy winter. Available
records led me to believe that in northern New England, during 1816, winter conditions were
drier and colder than normal; while farther south in Massachusetts the season was warmer and
wetter. Among the years with the greatest number of snowfalls were: 1792, 1804, 1805, 1806,
and 1818. Those years with the least number included: 1828, 1829 and 1834.

Reconstructions bearing evidence concerning the storminess of the period [e.g., the percentage
of days with westerly fair-weather winds (Figure 13) and the number of days in which thunder
and lightning storms occurred (Figure 14)], show that 1816 was rather stormy. The frequency
of westerly "fair weather bearing" winds was somewhat below the record mean; while days with
thunder and lightning storms numbered close to the record mean. Evidently there was a decrease
in storm activity during the 1790s. In the early 1800s, there was a small increase followed by
another decrease late in the decade. From 1 809 through 1 822, there was considerable year-to-year
variability but an overall increase in storminess. During the late 1820s and all of the 1830s, there
was a general decrease in storm frequency.

Figure 13: Westerly winds per year over southern New England, 1790-1839.

139

0'-wn^u^(CNeoo)o^wnttuitoNooc»o^cijt^^uiu)r^»aO'-(\iw^ui<fiNoooiO'-(\in^irtU3Ncoo)o
r^r^r^r^r^r^r^r^r^r^coaoaoaococooocDOOcocOQOoocoaoooaoaooocooocoaooocoooaoaoaococooocooocooococococDao

Year

Figure 14: Number of days per year with thunder and lightning storms for eastern Massachusetts, 1790-
1840.

Conclusions

1790 through 1839 featured two abnormally cold periods (1812 through 1818 and 1832 through
1838) and two warmer, relatively more stable periods (1799 through 1810 and 1819 through
1830). The warmth and stability of the latter two decades, compared with cold and relative
storminess of the 1810s, heightened peoples' awareness of the contrast between the two climatic
regimes. Especially in northern New England, where considerable farming took place on
climatically-marginal lands, the cold years brought disaster. To make matters worse, the swing
from warm to cold in the 1810s coincided with an increase in economic competition from the
midwestern United States and central Canada. The additional stress of crop shortfalls due to
shortened growing seasons forced many farmers to leave New England for what they believed
were more hospitable climates to the west (Smith et al 1981).

1816 was only one of several abnormal years that occurred during 1790 through 1839. When
viewed from this perspective, 1816's abnormality pales. Why then is 1816 so well remembered
while 1812 or 1836 are assigned to the second rank of the region's fabled years of climatic
adversity?

140

The answer lies not in our careful compilation of climatological records (for statistically 1816
does not measure up) but in the nature of 1816's abnormality and the impact of its greatly
shortened growing season on New Englanders' capability to raise food. A harsh, snowy winter
or severe spring flood have a great impact on certain segments of society, but a series of killing
frosts accompanied by a severe drought (especially in northern New England) hit nearly the entire
society by forcing up food prices for the rich and by reducing the available larder for the poor.
For the average New Englander, particularly the farmer, 1816 was the worst year of a series of
bad years. As time passed, 1816, in New England's folk memory, came to stand for the 1810s
as a whole. This idea has been passed down from generation to generation as the story of "the
year with no summer".

Acknowledgements and a Note on the Availability of Climatic Data

I thank the staff of the Historical Climate Records Office for their assistance. This research was
partially supported through Northern Arizona University's Organized Research Fund; I thank
particularly Henry O. Hooper, Associate Vice President for Academic Affairs, Research and
Graduate Studies for his support and assistance.

All databases discussed here are kept in the Historical Climate Records Office, part of the Center
for Colorado Plateau Studies at Northern Arizona University, Flagstaff. The Office has on file
and computer disc a large number of United States records collected for the seventeenth through
nineteenth centuries. There are particularly strong record groups for the northeastern United
States and the Colorado Plateau. The Office was founded by the author with the intention of
making these climatic materials available to other researchers. Record collection was done by the
members of the now disbanded Northeast Environmental Research Group centred at the
University of Maine and, after 1985, by the staff of the Historical Climate Records Office.
Record collection undertaken through 1985 was supported by grants from the National Science
Foundation and the Northeast Regional Experiment Stations to the University of Maine. A listing
of available records may be obtained by writing me.

References

Baron, W.R. 1988. Historical climates of the northeastern United States: seventeenth through
nineteenth centuries. In: Holocene Human Ecology in Northeastern North America. G.P.
Nicholas (ed.). Plenum Press, New York. pp. 29-46.

. 1989. Retrieving climate history: a bibliographical essay. Agricultural History 63. (in
press).

Baron, W.R., D.C. Smith, H.W. Borns, Jr., J. Fastook and A.E. Bridges. 1980. Long-Time
Series Temperature and Precipitation Records for Maine, 1808-1978. In: Life Sciences and
Agriculture Experiment Station Bulletin 111. University of Maine Press, Orono, Maine,
p. 97.

Baron, W.R., G.A. Gordon, H.W. Borns, Jr. and D.C. Smith. 1984. Frost-free season record
reconstruction for eastern Massachusetts 1733-1980. Journal of Climate and Applied
Meteorology 23:317-319.

141

Baron, W.R. and G.A. Gordon. 1985. A reconstruction of New England climate using historical
materials 1620-1980. In: Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus
55:229-245.

Cleaveland, P. 1867. Results of meteorological observations made at Brunswick, Maine, between
1807 and 1859, reduced and discussed by C.A. Schott. Smithsonian Contributions to
Knowledge 16:2-25.

Fichter, G.S. 1971. Eighteen-hundred-and-froze-to-death: snowfilled summer of 1816. Science
Digest 69(2):62-66.

Holyoke, E.A. 1883. A meteorological journal from the year 1786 to the year 1829, inclusive.
Memoirs of the American Academy of Arts and Sciences, New Series 1:107-216.

Hoyt, J.B. 1958. The cold summer of 1816. Annals of the Association of American Geographers
48:118-131.

Kirk, J.M. 1939. The Weather and Climate of Connecticut. State Geological and Natural History
Survey Bulletin 61 . Connecticut Geological and Natural History Survey, Hartford. 242 pp.

Landsberg, H.E. 1949. Climatic trends in the series of temperature observations at New Haven,
Connecticut. Geografiska Annaler 1(2): 125-132.

Landsberg, H.E., C.S. Yu and L. Huang. 1968. Preliminary reconstruction of a long time series
of climatic data for the eastern United States. Institute for Fluid Dynamics and Applied
Mathematics, University of Maryland, Technical Note B14-571 : 1-42.

Landsberg, H.E. and J.M. Albert. 1974. The summer of 1816 and volcanism. Weatherwise
27(4):63-66.

Leach, A. 1974. 1816 was a year for complaints. Barre-Montpelier Vermont, Times Argus,
22 July.

Milham, W.I. 1924. The year 1816 - the causes of abnormalities. Monthly Weather Review
52:563-570.

. 1950. Meteorology in Williams College. McClelland Press, Williamstown. 40 pp.

Mussey, B. and S.L. Vigilante. 1948. "Eighteen-hundred-and-froze-to-death". The cold summer
of 1816 and westward migration from New England. Bulletin of the New York Public
Library 1948:565.

Parsons, M. 1980. An eyewitness report of 1816, the cold year. Bittersweet 4-1:39-41.

Post, J.D. 1977. The Last Great Subsistence Crisis in the Western World. Johns Hopkins
University Press, Baltimore. 240 pp.

Reichmann, R. 1978. 1816 recalled as year without a summer. Providence, Rhode Island Sunday
Journal, 17 September.

142

Reiss, N.M., B.S. Groveman and CM. Scott. 1980. Seasonal mean temperatures for New
Brunswick, N.J. Bulletin of the New Jersey Academy of Science 1980:1-10.

Rodman, T.R. 1905. Monthly, annual and average temperatures and precipitation at New
Bedford, Mass., 1813-1904. Climate and Crops: New England Section, Annual Summary
1905:9.

Skeen, C.E. 1981. The year without a summer: a historical view. Journal of the Early Republic
1:51-67.

Smith, D.C., H.W. Borns, Jr., W.R. Baron and A.E. Bridges. 1981. Climatic stress and Maine
agriculture, 1785-1885. In: Climate and History. T.M.L. Wigley, M.J. Ingram and
G. Farmer (eds.). Cambridge University Press, Cambridge, pp. 450-464.

Stommel, H. and E. Stommel. 1979. The year without a summer. Scientific American 240:176-
186.

. 1983. Volcano Weather: The Story of 1816, The Year Without a Summer. Seven Seas

Press, Newport. 177 pp.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science
224(4654): 1191-1 198.

Wilson, CM. 1970. The year without a summer. American History Illustrated 5(6):24-29.

Wilson, H.F. 1967. The Hill Country of Northern New England: Its Social and Economic History
1790-1939. AMS Press, New York. 455 pp. (Reprint of 1936 edition).

Sources Directly Mentioned in 1816 Weather History

Bascom, R.H. 1816. Manuscript diary for Phillipstown, Massachusetts. American Antiquarian
Society, Worcester, Massachusetts.

Dickerson, M. 1816. Manuscript diary for Morristown and Trenton, New Jersey. New Jersey
Historical Society, Newark, New Jersey.

Featherstonhaugh, G.W. 1816. Manuscript diary for Albany, New York. Albany Institute,
Albany, New York.

Hill, B.T. (Editor). 1910. The diary of Isaiah Thomas 1805-1828. American Antiquarian
Society Transactions and Collections 10:251-263.

Hill, L. 1869. Meteorological and Chronological Register. L. Hill, Plymouth, Massachusetts,
pp. 61-66.

Kellogg, E. 1816. Manuscript notes in Old Farmer's Almanack for 1816. Maine Historical
Society, Portland, Maine.

143

Kimball, B. 1856. Extracts of a diary. In: The History of Concord, 1725-1853. N. Bouton (ed.).
B.W. Sanborn, Concord, New Hampshire, p. 771.

Lane, J. 1816. Manuscript diary for Sanbornton, New Hampshire. New Hampshire Historical
Society, Concord, New Hampshire.

Larcom, J. 1951. Diary of Jonathan Larcom of Beverly, Massachusetts. Essex Institute Historical
Collections 87:65-95.

Lincoln, T. 1816. Manuscript record book of weather reports. Maine Historical Society,
Portland, Maine, 1: (not paginated).

Longfellow, S. 1816. Manuscript diary for Gorham, Maine. Maine Historical Society, Portland,
Maine.

Miller, A. 1816. Manuscript diary for Wallingford, Vermont. Vermont Historical Society,
Montpellier, Vermont.

Robbins, B.F. 1816. Manuscript travel diary. Maine Historical Society, Portland, Maine.

Robbins, T. 1986. Diary of Thomas Robbins, DD, 1796-1854. Beacon Press, Boston, pp. 236-
247.

Sewall, H. 1816. Manuscript diary for Hallowell, Maine. Maine State Archives, Augusta, Maine.

Weare, J. 1912. The diary of Jeremiah Weare, York, Maine. New England Historical and
Genealogical Register 66:77-79.

Wheeler, W. 1930. Diary of William Wheeler. Yale University Press, New Haven, p. 221.

144

Extension of Toronto Temperature Time-Series from 1840 to 1778 Using
Various United States and Other Data

R.B. Crowe1

Abstract

Daily maximum and minimum temperatures for the city of Toronto are archived from 1 March
1840 to the present day. This lengthy time-series can be extended considerably by using standard
differences in mean monthly temperatures between Toronto and some United States stations, the
earliest of which dates from July 1778. In addition, there are considerable temperature data taken
three times a day from another station in Toronto in the 1830s. These data were adjusted and
monthly mean temperatures calculated. Mean July temperature for Toronto in 1816 is calculated
to have been remarkably low (nearly 16°C).

Introduction

In December 1839, the British Government established a meteorological and magnetic
observatory at Toronto, Ontario. Some sporadic observations began late in the month at Fort
York (on the shore of Lake Ontario just west of the town, then called York). Fixed hourly
observations of temperature commenced in the new year, but not until 1 March 1840 were
regular daily maximum and minimum values recorded: on this date daily Archive readings begin.
On 5 September 1840, the observation site was moved to the University of King's College (now
the University of Toronto) about 2 km north of the lakefront. Although a number of small
changes in location occurred in later years, a relatively homogenous, high-quality data set extends
from September 1840 to the present day. However, the rather large urban heat-island effect,
which influences the records of all such large cities, is evident.

The Toronto record from 1840 to the present comprises the longest continuous temperature time-
series in the Canadian climatological Archive, and thus it is frequently used in analyses of long-
term temperature trends.

The purpose of this paper is to present a method of extrapolating the Toronto temperature time-
series backwards from 1840, using various United States and other data. The earliest American
data used in the analysis were taken in 1778. The significance of these data for 1816 is
mentioned.

Sources of Early Climatic Data Used for Comparative Purposes
Toronto Area

The first fragmentary climatic data for Toronto were taken in the year 1801. These are contained
in the Hodgins Papers in the Archives of Ontario, Toronto, dating from the late nineteenth
century. The data are identical to those published in the Upper Canada Gazette at the time the
observations were taken, so presumably Dr. Hodgins merely copied long-hand this original

Canadian Climate Centre, Environment Canada, 4905 Dufferin Street, Downsview, Ontario M3H 5T4, Canada.

145

source. The data include the temperature and weather or sky condition at three fixed times a day.
Similar data were published for a number of months in the same newspaper around 1820. Neither
data set was lengthy enough to be used in this study.

Later a longer, more useful data set was taken by Dade (1831-41) from January 1831 to April
1841. Reverend Dade was the Headmaster of Upper Canada College, then situated close to the
centre of town on the lake shore, not far from the later Fort York station. The thermometer was
read two or three times a day at fixed times, but slight changes in reporting hours occurred
during the decade, and occasionally only one observation was taken in a day. Only a few months
were incomplete, except for an extended period from October 1838 to June 1839 when Dade
returned to England for the winter.

Periods of record for the various early Toronto stations used for comparative purposes are shown
(Figure 1). Data from Fort York are combined in the Archive with those from the University
station and identified as "Toronto" (no modifier), but is unofficially called "Toronto City". Data
later than 1855 were not considered.

EARLY EASTERN NORTH AMERICA
CLIMATIC DATA

1831-41
DADE

1840-70-
ORONTO

1827-60
.LOWVILLE

1835-45
ANCASTER

1831-50

LEWISTON

1831-70
BUFFALO

1830-63
• ROCHESTER

^1830-60
FREDONIA

1830-49 1826-63 1827-49
ONEIDA UT|CA#FA1RF|ELD

1827-49 • • 1827-60
# AUBURN ^HAMILTON

1832-63
• CORTLAND

1795-96
1813-14
1820-70_
£ ALBANY

1817-7<£? |
.BALTIMORE

1789-1795
1797
1804-15
1821-70
NEW YORK

1778-1870
NEW
• HAVEN

Figure 1: Early eastern North America climatic data.

146

Remainder of Southern Ontario

Data for Ancaster (about 65 km southwest of Toronto) were taken by Craigie (1835) from
January 1835 to December 1845, and proved to be of limited use in the Toronto data extension.
William Craigie was a surgeon who apparently tabulated daily maximum and minimum
temperatures as well as fixed-hour readings. His thermometers "were in a northern exposure, five
feet from the ground, and shaded from the effects of direct insolation and radiation from the
sky". However, only newspaper tabulations of monthly means of the 9 a.m. and 9 p.m.
observations survive.

American Stations

Mean monthly temperature data were abstracted from publications of the Smithsonian Institution
(1927) for Albany, New Haven and New York City and the United States Weather Bureau (1932-
37) for Albany, Baltimore and Rochester. Considerable monthly data were also available from
grammar schools in New York State (Hough 1855, 1872). Data for Auburn, Buffalo, Cortland,
Fairfield, Fredonia, Hamilton, Lewiston, Lowville, Oneida, Rochester (College) and Utica were
used, other stations listed in the above publications having insufficient useful data.

All stations actually used in the study are shown in Figure 2. Many months and years for most
stations were noted in the New York State grammar school records, and only the first and last
years of data are shown. In all cases, data later than 1855 were not used.

In the case of most of these early data, excepting Toronto (city), observations were taken with
the thermometer attached to the north wall of a building. Recording maximum and minimum
thermometers were not generally used. Monthly means were computed from two, three or more
observations a day, and the time and number of daily observations frequently changed and were
not consistent, either at a site or from one station to another. In addition, thermometers may not
have been calibrated accurately or sufficiently shielded from insolation, and changes in exposure
or siting may not have been recorded.

Method of Estimation of Toronto Mean Temperatures

Three distinct methods were employed in the calculation of Toronto mean temperatures due to
significant differences in the form of the source data: monthly means at the American stations
(calculated by a variety of methods depending on the station); daily data for Dade; and monthly
means for 9 a.m. and 9 p.m. in the case of Ancaster. These were labelled Method "S", Method
"D", and Method "A" (Figure 3).

All three methods were employed whenever data permitted. In deriving the final Toronto
estimates, however, Method "D" was chosen whenever Dade information was available. Thus,
Method "S" was used up to December 1830, but Method "D" from January 1831 to February
1840. For missing Dade months, Method "S" was substituted before 1835, but from this year on,
a linear regression equation was used based on the 52 months when the calculated American data
could be compared with both Dade and Ancaster calculations:

T = -0.126 + 0.6045 TA + 0.4108 Ts,

where T is the estimated Toronto monthly mean (°F),

TA is the estimated Toronto mean using Method "A",
and Ts is the estimated Toronto mean using Method "S".

147

J|F|M|A|M|J|J| A | S | O | N | D

I p T M I A I M I j I j I A 1 S 1 O 1 N 1 D

DADE (On Lakeshore)

FORT YORK

UNIVERSITY

Figure 2: Early Toronto climatological records.

(Mar.)1840-^ TORONTO

METHOD

"A"

METHOD

"D"

METHOD

"S"

1835-1845 ANCASTER

1831-1841

DADE

|~1778

FROM 1 TO 16 UNITED STATES STATIONS

1778-1830:

METHOD "S"

1831-1834:

METHOD "D"
(IF DADE MISSING: METHOD "S")

3. 1835- (Feb.)1 840:

METHOD "D"

(IF DADE MISSING: REGRESSION
EQUATION USING METHOD "S"
AND METHOD "A")

Figure 3: Data used to estimate mean monthly temperatures at Toronto.

149

Method "S"

This method essentially involved calculating standard differences between various American
stations and Toronto. For example, since August is normally 4.5°F (-15.3°C) cooler at Toronto
than at Albany, it was assumed that all missing August means at Toronto for which data were
available at Albany could be reasonably estimated by subtracting 4.5° from the Albany values,
no matter whether the month was near normal or significantly below or above normal.

Method "S" is outlined in Figure 4. There are five distinct steps:

1. Difference calculations. Mean monthly temperature data for Toronto (city), 1840 (March-
December) - 1870, and for all available American stations within about 400 miles (644 km)
of Toronto having significant data before 1870 were tabulated. Sixteen distinct United States
stations were available. Most stations did not have data before 1820, but New Haven had data
as early as 1778 (Figure 2). In order to facilitate comparison of data from all stations, for
each individual month, differences were calculated for all possible stations pairs, for example,
Toronto minus Rochester, Toronto minus Albany, Toronto minus New Haven, Rochester
minus Albany, etc.

2. Correction and deletion of bad data. The differences for each station pair were tabulated
by month and year, and the overall monthly-mean differences calculated. Also, standard
deviations of the mean monthly temperatures for each station were calculated. Then, for each
station pair for each of the 12 months, an average standard deviation (s.d.) was calculated,
and all differences greater or less than one standard deviation were identified. For example,
the s.d. of the August means at Toronto is 1.8°, Albany 2.2°, for an average of 2.0°. The
mean difference for the month, Toronto-Albany, is -4.5°, so that the 1 s.d. range is -2.5°
to -6.5°. By identifying differences outside the 1 s.d. range, it was easy to spot unusual
months or questionable data. For any month for which no station-pair differences lay outside
the 1 s.d range, the data were assumed to be reasonably good. Otherwise, a subjective
assessment was made by the areal plotting of the means and departures from normal. In a few
cases, it was possible to correct a value when it was obvious that a typographical error of
10°F (12.2°C) had been made in the printed source. Most questionable data, however, were
discarded.

3. Choosing of stations and periods for analysis. Following the corrections and the discarding
of questionable months, a new data set for each of the 16 American stations was prepared.
Toronto data were assumed to be "good", and the aim was to choose as many of the 16
United States stations as possible for the 1840-70 period for comparative purposes. Many
stations had missing or discarded data in the last half of this period, so it was necessary to
use only 1840-55. Within this period, not enough data were available for the computation of
reasonable monthly means at Auburn or Buffalo, and data for Rochester College were
identical to, or varied by only a small constant from, the Rochester data and hence was
suspect. The number of American stations useful for comparative purposes was therefore
reduced to 13.

4. Standard difference calculations. For each of the 12 months, mean differences in monthly
temperatures between Toronto and each of the 13 American stations were calculated. The
calculations were based on all months, September 1840 to December 1855, inclusive. I
considered that the early data for Toronto (March to August 1840), taken near the lake shore
at Fort York, were not homogeneous with the later University site observations. Because of

150

some missing months for most United States stations, standard differences were calculated
from means based on 10 to 15 years in most cases, and which varied from month-to-month
and from station-to-station. The standard differences between Baltimore and Toronto ranged
between 10.0° and 13.5°, depending upon the month. Because of its great distance from
Toronto, and the resulting high and variable differences in the monthly mean temperatures,
I decided to eliminate Baltimore. Thus, only 12 stations were left for further analysis.

5. Calculation of Toronto means. For each month, July 1778 to February 1840, an estimated
mean temperature for Toronto was calculated separately based on each American station for
which a mean monthly temperature was available. Thus, in June 1831, the mean monthly
temperature at Albany was 72.8°F (22.7°C), and since the standard difference, Toronto-
Albany (based on 1840-55) is -6.8°, the Toronto mean was estimated at 66.0°. Similarly,
New Haven was 71.1° in the same month, and the standard difference, Toronto-New Haven,
is -5.5°, so that Toronto's mean was calculated at 65.6°. The overall Toronto mean was
calculated as the unweighted average of all the individual estimates from the various
American stations, which for any month would vary from one to 12. This method of
estimating Toronto means was checked against the actual means from September 1840 to
December 1855. Although some monthly errors were as great as 4° F, the standard deviations
of the errors varied from 0.94 to 1.75 for individual months, averaging 1.17°.

Method "D"

In the case of Dade data, mean monthly temperatures had to be calculated from two or three
hourly observations a day. Then, standard difference calculations were made by comparing Dade
monthly means with those of several American stations. Toronto monthly means were also
compared to the same United States stations for the period 1840-55. In this way, a first
approximation was made of Toronto-Dade differences. The period of the Dade observations was
somewhat cooler on average than that of the later Toronto observations. Consequently, a second
approximation of the Toronto-Dade differences was made, allowing for mean temperature
differences between the two periods. When Toronto means were calculated month-by-month using
these second approximation differences, a comparison with the means obtained by using Method
"S" showed a consistent low bias. Hence, a final approximation of Toronto-Dade differences, and
therefore, calculation of Toronto means, was made to allow for this low bias.

Method "D" is outlined in Figure 5. There are eleven distinct steps:

I. Estimation of daily maximum and minimum temperatures. When Reverend Dade began
observations on 1 January 1831, he took readings at 9 a.m., 3 p.m and 6 p.m. However, this
pattern did not continue in later months and years. Sometimes the morning reading was at
7 or 8 a.m., but the mid-day observation was usually taken at noon and the evening one at
5 p.m. To complicate matters, while there was always a morning reading (excepting the
months with partial days, which were excluded from analysis), sometimes there was in
addition only a mid-day reading, sometimes only an evening one, sometimes both and
sometimes neither. It was necessary, therefore, to estimate daily maximum and minimum
readings. This was done by using mean hourly temperatures in comparison with mean daily
maxima and minima for each month at Toronto's Pearson International Airport (Atmospheric
Environment Service 1978). Thus, a correction factor was calculated to be subtracted from
the morning reading to estimate the daily minimum and to be added to the mid-day and
evening observations to estimate the daily maximum. These applied to Pearson Airport, so
corrections for Dade were computed by multiplying the Pearson corrections by the ratio of

151

U.S. MONTHLY
DATA

1 to 16 U.S. Stations 1778-1870
and Toronto 1840-1870

DIFFERENCE
CALCULATIONS

For each month all possible station pairs
(Stn. A - Stn. B, Stn. A - Stn. C,
Stn. B - Stn. C, etc.

CHOOSE STATION
AND PERIODS FOR
ANALYSIS
PURPOSES

STANDARD
DIFFERENCE
CALCULATIONS

CALCULATION
OF MEAN
TEMPERATURE AT
TORONTO USING
U.S. STATIONS DATA
AND STANDARD
DIFFERENCES

For each of the 12 months
Sep. 1840 - Dec. 1855 mean
differences in monthly
temperatures between Toronto
and each of 12 U.S. stations

For each month July 1778
Feb. 1840 calculated
separately for each
station and unweighted
average taken

Figure 4: Method "S'

152

the monthly mean daily range at Toronto (city) to that at Pearson. These figures were
rounded to the nearest whole Fahrenheit degree. According to modern practices, the
"climatological day" for a climatological station that takes only observations in the morning
and evening ends with the morning observation as far as the daily maximum for the previous
day is concerned. The daily minimum for Dade observations was calculated as the lowest of:
(a) the actual evening reading the day before; (b) the morning observation corrected for the
diurnal minimum; (c) the mid-day actual reading; and (d) the evening actual reading.
Similarly, the daily maximum for Dade was calculated as the highest of: (a) the morning
actual reading; (b) the mid-day observation corrected for the diurnal maximum; (c) the
evening observation corrected for the diurnal maximum; and (d) the morning actual reading
the following day. In those rare cases where neither mid-day nor evening observations were
taken, the minimum temperature for the day was calculated as above. Then a maximum was
estimated using the mean daily range for the month at Toronto (city). This value was checked
against the morning temperature the following day, and the higher of the two taken. Hence,
the daily maximum and minimum calculations took under consideration abnormal diurnal
temperature trends.

Calculation of mean monthly temperatures. For each month, the mean daily maximum and
mean daily minimum were calculated from the daily values. The mean monthly temperature
was then simply the mean of the mean daily maximum and the mean daily minimum.

Standard difference calculations, Dade minus United States stations. In order to estimate
Toronto mean temperatures by using Dade data, it was necessary to compare Dade monthly
means as computed above to those of as many American stations as possible. For each of the
12 months for the period January 1831 to April 1841, mean differences in monthly
temperature were calculated between Dade and each of nine American stations: Albany,
Cortland, Fredonia, Lewiston, New Haven, New York, Oneida, Rochester, and Utica. Data
for Fairfield, Hamilton and Lowville were not used in this analysis due to many missing
months of information during the decade.

Standard difference calculations, Toronto minus United States stations. Similarly, for
each of the 12 months for the period March 1840 to December 1855, mean differences in
monthly temperatures were calculated between Toronto and each of the nine American
stations used in the Dade standard differences above.

First approximation of Toronto-Dade differences. Since both Toronto and Dade means are
compared to the same nine American stations, the first approximation of Toronto-Dade
differences was obtained by subtracting the Toronto standard differences above from the Dade
standard differences above. These were calculated separately for each of the nine American
stations, and the overall mean taken for each month. This analysis indicated that, for every
month of the year, Dade values were high, and that correction values ranging from -0.6°
(February) to -4.0° (July) had to be applied to Dade means to give a reasonable estimate of
Toronto means. For the bitterly cold December of 1831, as an example, the Dade calculated
mean was 15.8°. Since the correction value for December is -2.2°, the first approximation
of the Toronto mean for the month would be 13.6°F (10.2°C).

Comparison of mean temperatures for 1831-41 with 1840-55. There was no reason to
assume that the whole period 1831-55 was climatologically homogeneous. In order to obtain
a measure of the differences in mean temperature between the early period of the Dade

153

observations (1831-41) and the later period of the Toronto observations (1840-55),
calculations were performed for the three United States stations with the best and most
continuous observations, Albany, New Haven and New York. Means were calculated for
each month separately for each of the three stations and for both periods, allowing for those
months when Dade observations were missing. For each month, an overall mean difference
(unweighted average of the three stations) between the period means was obtained. The
earlier period was colder than the later at each of the three stations for each of the 12
months. The overall monthly differences ranged from 0.4°F (-17.6°C) for April to 2.5°F
(-16.4°C) for December.

7. Second approximation of Toronto-Da de differences. The second approximation considers
the fact that the earlier 1831-41 period was significantly colder than the later 1840-55 period.
The first approximation Toronto-Dade difference in mean temperature is -4.0°. The
difference between the two periods for the same month is 1.2°, so that the total correction
applied to Dade means to obtain Toronto means is -5.2°F. Because of the variability from
month to month, Fourier smoothing was applied to the monthly values. As a result, the
second approximation of Toronto-Dade differences ranged from -2.7° in September and
October to -4.1° in December. Again, in the case of the frigid December of 1831, the Dade
mean of 15.8° with a correction of -4.1° results in a Toronto mean of 11.7°F (-11.3°C).
This is 1.9° lower than the first approximation calculation.

8. Preliminary calculation of mean Toronto temperatures using Dade and second
approximation differences. For each month for which Dade means were available in the
period January 1831 to April 1841, a Toronto mean was calculated using the second
approximation differences (Fourier-smoothed) above.

9. Comparison of mean temperatures at Toronto by using Dade calculations above and by
using Method "S". For each month for which Dade means were available in the period
January 1831 to April 1841, the Toronto mean using the Fourier-smoothed second
approximation differences with Dade were compared with means as calculated by Method "S"
(using all available American data). The overall mean differences in the two methods were
compiled for each month and it was found that Method "S" gave higher values than the Dade
method in all months - ranging from 0.4° in March to 2.6° in February. The standard
deviation of the monthly differences between the two methods ranged from 0.6° in June to
1.7° in January. In the case of the frigid December of 1831, Method "S" indicated a Toronto
mean of 14.1°F (-9.9°C), 2.4° higher than the preliminary calculation using Dade.

10. Final approximation of Toronto-Dade differences. Since Method "S" indicated somewhat
higher means for Toronto for every month of the year than those by using the preliminary
Dade calculations, apparently the second approximation allowing for the mean temperature
differences between the two periods was based on differences that were too great.
Consequently, the final approximation of Toronto-Dade differences was calculated by
reducing the second approximation differences by the differences indicated between Method
"S" and the preliminary Dade calculations above. Thus, the December second
approximation Toronto-Dade differences, Fourier-smoothed, is -4. 1 °, the correction due to
the Method "S" comparison is + 1.7°, so the final Toronto-Dade correction for the month
is -2.4°. Again, because of the month-to-month variation in the correction values, a Fourier
smoothing was applied. The final approximation of Toronto-Dade differences then ranged
from -1.1°F in October to -3.2°F in June. In the case of bitterly cold December 1831, the

154

smoothed final Toronto-Dade difference is -1.9°, so that when applied to the Dade mean
of 15.8°, the Toronto estimate works out to be 13.9°, very close to the 14.1°F (-9.9°C)
in the case of Method "S".

1 1 . Final calculation of Toronto means. For each month for which Dade means were available
in the period January 1831 to February 1840, a Toronto mean was calculated using the final
approximation differences (Fourier-smoothed) above.

Method "A"

In the case of Ancaster data, published monthly means based on two observations per day, 9 a.m.
and 9 p.m., had to be corrected to standard monthly means based on daily maxima and minima.
Then, standard difference calculations were made between the overlapping records of Ancaster
and Toronto . From these, estimates were made of Toronto means for those months before
records began in March 1840.

Method "A" is outlined in Figure 6. There are three distinct steps:

1. Correction of mean monthly temperatures. No daily observations are available for
Ancaster, only published monthly means of 9 a.m. and 9 p.m. observations, from which a
simple average was computed to produce a monthly mean. Correction values were calculated
for each month in order to provide monthly means based on the modern practice of using
daily maxima and minima. These were done by comparing means produced by averaging 9
a.m. and 9 p.m. monthly means at Toronto's Pearson International Airport (Atmospheric
Environment Service 1978) with the monthly means at the same station, which are calculated
by the usual mean of daily maximum and minimum values.

2. Standard difference calculations. Ancaster and Toronto data overlap for the period March
1840 to December 1845. For each of the 12 months during this period, mean differences
were calculated between the corrected Ancaster monthly means and the official Toronto
means.

3. Calculation of Toronto means. For each month, January 1835 to February 1840, a mean
temperature was calculated for Toronto using the corrected Ancaster mean and the standard
differences above.

The Reconstructed Toronto Temperature Time-Series

By using Methods "S", "D" or "A" as appropriate, monthly mean temperatures were estimated
for Toronto from July 1778 to February 1840. No American data were available for September
1778 or February, July, August, October, November and December 1779, so that no means
could be estimated for these months. Beginning with January 1780, a complete set of monthly
values was obtained. All calculations were done using the Fahrenheit scale, and then the whole
set was converted to Celsius and combined with Atmospheric Environment Service Archive values
that begin March 1840 and continue with no breaks until the present day.

Statistical F tests were applied to monthly and seasonal values to test the homogeneity of variance
between various periods. In the first instance, three 30-year periods were chosen: (A) 1780-1809;
(B) 1810-39; and (C) 1840-69. Period A involves only Method "S": through much of this period

155

Original daily data Jan. 1831 - Apr. 1841
(some months missing) variable hours
usually 2 or 3 times per day

CALCULATION OF DADE

DAILY MAXIMUM AND
MINIMUM TEMPERATURE

CALCULATION OF
DADE MEAN MONTHLY
TEMPERATURES

STANDARD DIFFERENCE
CALCULATIONS DADE
MINUS U.S. STATIONS

STANDARD DIFFERENCE
CALCULATIONS TORONTO
MINUS U.S. STATIONS

FIRST APPROXIMATION
OFTORONTO-DADE
DIFFERENCES

Based on mean daily temperature
cycles for the Toronto area and
modern "cl imatological day" practices

Based on average of daily maxima
and minima for all complete months

For each of the 12 months (Jan. 1831 - Apr. 1841)
mean differences in monthly temperatures between
Dade (D) and each of 9 U.S. stations
(S) (Y = D - S)

For each of the 12 months (Mar. 1840 - Dec. 1855)
mean differences in monthly temperatures between
Toronto (T) and each of the same 9 U.S. stations
(S) (X = T - S)

For each of the 12 months mean differences in
monthly temperatures between Toronto and Dade
assuming "S" is the same for both periods

(Z=x-y=T-D)

Figure 5: Method "D".

156

Figure 5: (cont'd)

METHOD "D" - CONTINUED

CFROM \
PREVIOUS SHEET )

COMPARISON OF MEAN
TEMPERATURES PERIODS 1831-
1841 WITH 1840-1855

SECOND APPROXIMATION
OFTORONTO-DADE
DIFFERENCES

PRELIMINARY CALCULATION OF
MEAN TEMPERATURE AT
TORONTO USING DADE AND

SECOND APPROXIMATION DIFFERENCES

COMPARISON OF MEAN TEMPER-
ATURE AT TORONTO USING DADE
ABOVE AND BY USING METHOD "S"

For each of the 12 months mean differences (C)
in monthly mean temperatures at 3 U.S. stations
Period A(1831-1841) from Period B(1840-1855)
(C=B-A)

For each of the 12 months difference C
substracted from first approximation differences
(Z, = Z - C) (Zj 12-month Fourier smoothed)

For each month, Jan. 1831 - Apr. 1841, Toronto
mean (Tq) calculated using Dade mean (D) and
Fourier smoothed 2nd approximation difference
(ZJ (TD = D + ZJ

For each month Jan. 1831 - Apr. 1841
difference calculated using Dade (Tn)
from Method "S" (Ts) (C, = Ts - Tp)

FINAL APPROXIMATION
OFTORONTO-DADE
DIFFERENCES

FINAL CALCULATION OF MEAN
TEMPERATURE AT TORONTO
USING DADE AND FINAL
APPROXIMATION DIFFERENCES

^OUTPUT^

For each of the 12 months difference C,

substracted from Fourier smoothed lx

(Z2 = Zj - Cj) (Z2 12-month Fourier smoothed)

For each month Jan. 1831 - Feb. 1840 Toronto
mean (Tp) calculated using Dade mean (D)
and Fourier smoothed final approximation
differences (Z2) (Tp = D + Z2)

157

Published monthly means Jan. 1835 - Dec. 1845
based on means of two observations per day.
9 a.m. and 9 p.m.

CORRECTION OF
MEAN MONTHLY
TEMPERATURES

Means based on 9 a.m. and 9 p.m. converted
to estimated means based on daily maxima
and daily minima using mean daily temperature
cycles for the Toronto area

STANDARD
DIFFERENCE
CALCULATIONS

For each of the 12 months, Mar. 1840
Dec. 1845, mean differences in monthly
temperatures between Toronto and Ancaster

CALCULATION OF
MEAN TEMPERATURE AT
TORONTO USING ANCASTER
AND STANDARD DIFFERENCES

(^OUTPUT^

Figure 6: Method "A".

For each month Jan. 1835 - Feb. 1840

158

only one, two or three American stations had data available for comparative purposes. Period B
involves Dade data as well as an increasing number of United States stations. Period C involves
the early instrumental record at Toronto before urban warming was significant. Only June
temperature variances are significantly different at the 99% level between periods A and B and
between A and C. In the second instance, two 40-year periods were chosen: (A) 1801-40 and
(B) 1841-80. Period A represents the last 40 years of reconstruction prior to the official
observations beginning in March 1840, whereas Period B contains the first full 40 years of
instrumental data. Only January variances are significantly different at the 99% level.

25°C

Figure 7: Mean July temperatures at Toronto.

159

Individual plotted mean monthly July temperatures from 1780-1870 are shown (Figure 7). The
cold July of 1816 (the year without a summer) is immediately evident. The trend line is based
upon a 50-year running mean.

OOO O OOOOOO OOOOOO

Y- CM CO TT IT) CD h- CO CD O 1- CM CO *T lO IT)

CO CO CO CO 00 CO CO CO CO C7> CD CD CD CD CD CD

Figure 8: Trend lines based on 50-year running means of mean monthly and annual temperatures at
Toronto (values before 1804 and after 1962 are considered constant).

160

In Figure 8 trend lines based upon 50-year running means plotted for the middle year are shown
for all months. Fifty-year means for 1780-1987 can be plotted only from 1805 to 1962.
Temperatures were considered constant before and after these dates. The overall rise in
temperature during the period ranges from 0.9°C in January to 3.3°C in October, with an annual
average of 2.2°C. A significant amount of this increase is no doubt due to the urban heat-island
effect, which became increasingly significant from the 1880s on.

References

Atmospheric Environment Service. 1978. Hourly Data Summaries - No. 3R, Toronto
International Airport, Ontario. Climatological Services Division, Atmospheric
Environment Service, Downsview, Ontario, p. 28.

Craigie, W. 1835. Mean results for each month of eleven years (1835 to 1845, inclusive) of a
Register of the Thermometer and Barometer, kept at Ancaster, C.W. Clippings from
newspapers, names and dates unknown. (Unpublished manuscript, Atmospheric
Environment Service, Downsview, Ontario).

Dade, Reverend C. 1831-41. Temperatures at Toronto. 47 pp. (Unpublished manuscript,
Atmospheric Environment Service, Downsview, Ontario).

Hough, F.B. 1855. Results of a series of Meteorological Observations, Made in Obedience to
Instructions from the Regents of the University, at Sundry Academies in the State of New
York, from 1826 to 1850, Inclusive. Weed, Parsons and Company, Albany. 502 pp.

. 1872. Results of a Series of Meteorological Observations, Made under Instructions from

the Regents of the University, at Sundry Stations in the State of New York, Second Series,
From 1850 to 1863, Inclusive. Weed, Parsons and Company, Albany. 406 pp.

Smithsonian Institution. 1927. World Weather Records. Smithsonian Miscellaneous Collections
79. The Lord Baltimore Press, Baltimore. 1199 pp.

United States Weather Bureau. 1932-37. Climatic Summary of the United States; Climatic Data
Herein from the Establishment of Stations to 1930, Inclusive. Third Edition. United States
Government Printing Office, Washington, D.C.

161

Climate in Canada, 1809-20: Three Approaches to the Hudson's Bay
Company Archives as an Historical Database

Cynthia Wilson1
Abstract

The Hudson's Bay Company archives are rich in weather information. Material includes:
Meteorological Registers; descriptive entries encapsulating the day's weather or seasonal comment
in the Post Journals, Correspondence and Annual Reports; and proxy weather data. As a database
for studying past climate, the strength of the archive is its diversity, permitting cross-checking
of results and the convergence of evidence. But the problem of fragmentary evidence has to be
overcome.

This paper briefly describes three approaches I have taken to integrate the different material, in
studying May-October climate during the nineteenth century along the east coast of Hudson/James
Bay, and over east/central Canada: (1) to establish a detailed regional climatology (1814-21) as
an historical benchmark; (2) to obtain year-by-year (1800-1900) estimates of monthly temperature
anomalies (reference 1941-70) and wetness indices; (3) to produce schematic daily weather maps
(east/central Canada, 1816-18).

The results of these studies indicate: (1) from 1800-10, May-October temperatures along the east
coast of the Bay were akin to those today. The seasons then became cooler, with mean
temperatures falling spectacularly in 1816 and 1817 to values below the modern record; from
181 1-20, they averaged about 1.6°C below the 1941-70 normal; (2) the low temperatures were
accompanied by greater-than-normal snowfall, and in 1816 and 1817, they essentially precluded
the growth of many plants. Some areas of snowcover probably remained through the season in
1816, and offshore, the Bay was barely free of ice at the end of the season; (3) flow patterns over
east/central Canada from 1 June to mid-July 1816 suggest that spring had been delayed or
protracted by as much as six weeks; (4) although there was some recovery in 1818, seasonal
temperatures below the 1941-70 energy level remained characteristic until the 1870s.

The period of unrelieved record cold from October 1815 to March 1818 may have been
influenced by Tambora - a relatively short-term volcanic eruption exacerbating a longer-term (60-
year) lowering of temperature already underway. But there can be no doubt that the exceptionally
heavy Bay ice lingering through the summers combined with the high frequency of onshore north
and west winds was a major factor in reducing summer temperatures on the east coast of the Bay.

Introduction

Until well into this century, climate loomed large in the daily living and even survival of the
those inhabiting the Canadian Shield and Prairies, and the Hudson's Bay Company (HBC) Post
Journals, Correspondence and Annual Reports provide a rich variety of information directly or
indirectly pertaining to weather.

90 Holmside, Gillingham, Kent ME7 4BE, U.K.

162

The climatic information is of three kinds: (1) Meteorological Registers, often meticulously kept
in accordance with the accepted practices of the time, but with one or two important exceptions
on the east side of Hudson Bay, the periods of record are relatively short; (2) descriptive entries
in the Post Journals encapsulating the day's weather, with occasional seasonal comment (the latter
is also found in the Correspondence and Annual Reports); (3) proxy weather information, the
impact of weather on the natural environment, and on the property, activities and well-being of
the inhabitants (in all three sources).

In using this remarkable record as a database to extend the modern climatic series into the past,
and to study past climatic anomalies, a major problem is that of fragmentation. This has resulted
from accidents of history, Company policies and activities, and from the nature and interests of
the individuals recording the events. The strength of the archive is its diversity, permitting cross-
checking of results and the convergence of evidence. This paper describes briefly three
approaches that I have used to integrate the material, so as to overcome the fragmentation and
take full advantage of the diversity in reconstructing (Figure 1):

1. A regional climatology1, as a climatic benchmark in the historical record. (The east coast of
Hudson/James Bay, 1814-21).

2. Extended seasonal time series1 - temperature and wetness indices. (The east coast of
Hudson/James Bay, 1800-1900).

3. Schematic daily weather maps2. (East/central Canada, summers 1816-18. This study is still
in its early stages).

Some aspects of the reconstructed climate in Canada from 1809 to 1820 have been selected to
illustrate the rich potential of the HBC archives as an historical database.

The overall approach to the historical material was traditional, in which the researcher does the
abstracting so as not to lose vital information offered by the context and subtext. With this in
mind, weather and proxy data were abstracted in context. Climatically, the historical data were
approached as far as possible in physical terms, from the standpoint of small-scale climatology,
the approach of Landsberg (1967) and Geiger (1965). Even with the synoptic mapping, this
approach was helpful in evaluating and interpreting the individual point data. In this, personal
experience of several summers in the field at Great Whale3, observing the weather, measuring
surface energy exchanges and keeping weather journals, has played an integral part.

Owing to the detailed nature of this kind of work, I do not have space here to substantiate the
methods or to discuss the assumptions and confidence limits. This information is available,
together with a full account of the results, in four reports distributed by the Canadian Climate
Centre (Wilson 1982, 1983a, 1985a, 1988) and three papers published by the National Museum
of Natural Sciences in Syllogeus (Wilson 1983b, 1985b, 1985c); see also Wilson 1985d.

Studies 1 and 2 were carried out under contract to the Canadian Climate Centre, Atmospheric Environment Service,
Downsview, Ontario.

The developmental stages of study 3 have been funded by the National Museum of Natural Sciences, Ottawa, as
part of the Museum's Climatic Change in Canada Project.

I am grateful to the Centre d 'Etudes nordiques, Universite Laval, Quebec, the National Research Council of Canada
and the Canadian Atmospheric Environment Service for the logistical support and research funds which made this
possible.

163

STUDIES A.B. EAST COAST HUDSON/JAMES BAY : MAY TO OCTOBER

(LITTLE WHALE RIVER, GREAT WHALE, FORT GEORGE, EASTMAIN )

A 1814-1821

CLIMATOLOGICAL STUDY
AS A CLIMATIC BENCHMARK

1900

1916

NORMAL REFERENCE
PERIOD: 1941-70

1986

B. 1800 TO 1900

MEAN MAY -TO -OCTOBER TEMPERATURE
SERIES a WETNESS INDICES

MODERN METEOROLOGICAL RECORDS

EAST /CENTRAL CANADA
MAY TO AUGUST 1816-1818
SCHEMATIC DAILY WEATHER
MAPS

POSTS WITH JOURNALS FOR
• AT LEAST PART OF PERIOD
MAY -AUG. 1816, 1817, 1818

_ NON-HBC SOURCES FOR
SAME PERIOD

100W

Figure 1: Studies of past climate in Canada: three approaches to the Hudson's Bay Company archives as
an historical database.

Place and Time of Study

Considering the amount of work involved, the choice of region and period for study is critical.
For the regional climatology and the construction of the time series, the east coast of
Hudson/James Bay (HBC Posts: Eastmain, Fort George, Great Whale, Little Whale River) and
the active season (May-October) were selected for reasons that follow.

164

The Hudson Bay region appears to be particularly sensitive to climatic fluctuations, and the
eastern windward coast of the Bay provides an excellent laboratory. It is a marginal area with
respect to the fluctuating arctic/subarctic boundary and the northern limit of tree growth. This
vast sea, with its seasonal ice cover, open to arctic waters and arctic ice, extends the influence
of polar climate into the heart of the continent (south of Latitude 52 °N) in spring, and remains
a cold sink in summer; in late autumn the presence of open water creates a snowbelt on this
windward east coast. To the east lies the plateau of New Quebec/Labrador, a former centre of
the Laurentian Ice Sheet. All forms of life are so finely tuned to climate along these marginal
coastlands that any unusually severe or prolonged anomaly can soon disturb the ecological
balance, and human life and activity - and incidentally make good copy for Journal writers. Other
reasons for the choice of region include the availability of adequate modern weather and
environmental records, and my first-hand knowledge of the area.

The earliest HBC Post Journal for this region is for Eastmain in 1736, but I began the time series
with the nineteenth century, when better coverage was available for the warm season. With few
exceptions, there was at least one Post reporting through each season during the century. In 1814,
the Hudson's Bay Company gave top priority to a carefully defined program of weather and
weather-related observations at their Posts in Canada. With wars closing markets in America and
Europe, fiery competition in the field from the North West Company and a worsening economic
and social climate at home, the aim was to study and develop the local agricultural potential at
each Post, to cut the high costs of sending out European food. At a number of Posts, including
Great Whale, Fort George and Eastmain, fixed-hour temperature and weather data were also
recorded regularly in Meteorological Registers. The directive fell into abeyance after the
amalgamation with the North West Company in 1821. Although the Company may never have
applied its hard-won information, the archives from 1814 to 1821 remain a rich data source for
detailed regional climatological studies of a period of unusual climatic interest, and were used
here.

Again, by implementing this programme in 1814, the Hudson's Bay Company through its
network of Posts and lines of communication in central, western and northern Canada provided
a system of synoptic weather observation unique at this time, both in the discipline imposed, the
consistency of purpose and of manner of observing and recording, and in the extent of its
coverage. These Company records, together with the logs of the annual supply ships from
England, coastal shipping and of canoe journeys, offer a basis for synoptic weather mapping.

From the results of the climatological study for the east coast of the Bay, and given the interest
in the atmospheric circulation in the years around the eruption of Mount Tambora, the summers
1816 to 1818 were chosen for initial mapping and investigation. Figure 1 locates all HBC Posts
with Journals for at least part of the period May-August 1816-18. The density of the network and
type of weather information available varies through the season and from summer-to-summer,
depending in part on the regular seasonal operations and needs of the fur trade itself, but to a
greater extent on the conflict between the Hudson's Bay and the North West companies. Sadly,
the battle for the Athabasca trade curtailed weather information from the Red River Valley
westward from June 1816 through 1817. For the three summers, no alternative historical weather
sources have been found for the west. For eastern Canada and northeastern United States a
number of weather records, personal diaries, Mission reports and newspaper articles are available
for this period to extend coverage (see Figure 1). Regular weather observations were also
recorded at Godthaab (now Nuuk), Greenland. The search for additional information continues.

165

The Three Approaches

A Regional Climatology, 1814-21

The prime weather data sources in the HBC archives are the Meteorological Registers, even
where they were kept for only a few years. To try to make full use of them, one approach is to
analyze and integrate all aspects of the local or regional weather from all HBC sources for those
years - temperature, winds, cloud, precipitation, extreme events and, rarely, pressure, together
with all proxy-weather indicators - to gain as clear a picture as possible of the climate of that time
in terms of the present-day climate: that is, to set up a climatic benchmark.

The dangers of this approach are only too well-known: the differences between the historical and
modern instrumentation, exposure, observing practices and so on. Happily, a tradition of careful
meteorological observation and reporting had become established on the west side of the Bay in
the second half of the eighteenth century with the collaboration of the Royal Society, which
advised on instruments and procedures. This tradition continued into the early nineteenth century.
Provided that basic assumptions are made explicit and their import is clearly stated, and that
every effort is made to compare like with like, I believe the results to be worth the time and
effort required.

With the historical temperature readings at Great Whale/Fort George and Eastmain, I tackled the
calibration from several directions, hoping in this way to approach a consensus and to avoid
circular arguments. The three main lines of attack were:

1. Historical - the reconstruction of the early observing sites, and social context, and of the
meteorological instrumentation and procedures accepted at the time. A study was also made
of the history and. homogeneity of the respective modern temperature records.

2. Physical - examining the systematic temperature differences that might arise from changes
in site, instruments and their exposure, and observing practices, given the distinctive qualities
of the subarctic surface conditions and regional and local weather.

3. Statistical - an application and extension of current Canadian quality-control procedures, the
fitting of simple regression models, and the analysis of the fields of error.

Although corrections were made to the daily maximum and noon temperatures, and also to the
daily minimum where the start of the climatological day differed, I was impressed by the
consistency of the historical temperature record in the context of the instruments and procedures
of the time.

Extended Seasonal Time Series, 1800-1900

The basic HBC sources of continuous and consistent weather information over extended periods
are the descriptive entries in the Post Journals. A reading of the Journals often leaves a strong
impression as to the relative heat or cold, dryness or wetness of the different seasons, through
the subtle integration of the many different weather and environmental factors and their impact.
Thus a second approach is to try to integrate daily and seasonal weather remarks and all forms
of proxy-weather information, to obtain monthly estimates of the temperature anomaly with
respect to a modern reference period (in this case 1941-70); also, to obtain monthly wetness
indices with respect to the modern precipitation record. The benchmark set up in the first study
acts as a useful reference for the early part of the century.

166

The approach is similar to that taken by Pfister (1980) in Switzerland. For the thermal series, a
first approximation to the monthly temperature anomaly is obtained from the direct weather
remarks, then the proxy indicators are applied to try to obtain an order of magnitude (timing,
intensity, duration), or at least supporting or modifying evidence. One major difference here is
that a greater variety of information must be used to compensate for the fragmentation of
individual data series, to give a convergence of evidence.

As the method must accommodate so many different kinds, fragments and combinations of data,
some firm climatological structure is required. A secure, yet flexible modern frame of reference
was provided by the nine modern daily temperature curves for Great Whale, Fort George and
Eastmain, respectively - comprising for each day of the May-October season the reference period
daily mean, the highest and lowest daily mean, the daily mean maximum and the highest and
lowest maximum, the daily mean minimum and the highest and lowest minimum. Other aids
included a wide variety of modern temperature and temperature-related information, including
analogues for warm and cold months.

The proxy-weather sources (Figure 2) can be grouped into three: snow and ice, phenological
events (plants and animals), and human activities. These data have been approached from two
points of view:

MAY

MID-SUMMER

GARDENING. POTATOES, TURNIPS, CABBAGES, BARLEY, OATS ETC.

OCTOBER

DIGGING/
MANURING

PLANTING / TRANSPLANTI NG
EMERGENCE HOEING

HARVESTING

o°c

LATE FROSTS

PLANT PESTS
"THE GRUB"

EARLY FROSTS

5°C

10 C

10 c

BREAK-UP
(RIVER)

SNOW/ S LE ET/ R Al NFALL

FROZEN GROUND
SNOWCOVER

I CLEARING
DRIFTS

WHITE PARTRIDGES/GEESE

FARM ANIMALS OUT
I TO PASTURE
OTHER ANIMALS
HAULING I

TRANSPORT ICE /WATER

NATURAL VEGETATION
BUDDING AND LEAFING
OF DECIDUOUS BUSHES

GROWTH OF GRASS

o°c

WILD FRUITS

CHANGE IN COLOUR
AND FALL OF LEAVES

FREEZE-UP
(RIVER)

HAYMAKING

BITING FLIES

BAY ICE

WHALES

RAI NFALL / SLEET/ SNOW
I

SNOWCOVER

GEESE/WHITE PARTRIDGES
FARM ANIMALS
I HOUSED
OTHER ANIMALS
I HAULING
TRANSPORT WATER/ ICE

SPRING

AUTUMN

Figure 2: Summary of environmental indicators.

167

1 . To obtain an indication of the timing and magnitude of seasonal and unseasonable events, and
to compare where they intersect the modern daily or weekly temperature curves. One avenue
of attack is the heat-unit concept, with thresholds 0°C, 5°C and 10°C; another, the cardinal
points with respect to the different crops; a third, the agroclimatic capability classes, which
delimit in climatic terms the crop potential of the region - and so on.

2. To try for a statistical link between the proxy data and the monthly temperature anomaly
during the modern period of record, to provide guidelines or rules of thumb.

All the information was integrated for each season and each Post in direct comparison with the
modern reference period to give the temperature anomaly month-by-month from May to October.
Where more than one Post was reporting, the anomalies were then compared and combined. The
period of greatest confidence is 1815-20.

The wetness index is a five-point scale based on the number of days with reported precipitation,
together with supporting remarks and proxy evidence indicating its intensity or duration, and the
degree and duration of dry periods. Each month and each season, for each Post, was assessed
directly against the modern record of precipitation at the respective weather stations. Following
Pfister, the upper and lower quartiles were chosen as limits between wet and dry months; also,
the octiles define very wet and very dry classes. Adjustments were made to account for those
days when rain fell only at night and may not have been reported, and to account for discrepancy
in the modern reporting of the number of days with snowfall between 24-hourly observing
stations and climatological stations (cf. Ashmore 1952; Manley 1978).

Schematic Daily Weather Maps, Summers 1816-18

A third approach permitting the integration of all available kinds and fragments of weather
information is synoptic mapping, although the Meteorological Registers with their detailed timed
weather observations act as linch pins in the historical analysis.

Since the historical data include few records of atmospheric pressure at this time, these weather
maps for east/central Canada are primarily based on surface wind data - more readily and
frequently recorded. Thus they offer schematic representations of the flow patterns, rather than
the refinement of Kington's (1988; this volume) classic series of daily weather maps for western
Europe and the northeastern Atlantic in the 1780s, and in 1816, which are firmly based on a
network of barometer readings. That surface winds can be so used has already been elegantly
demonstrated by Lamb (in association with Douglas) for the period of the Spanish Armada, May-
October 1588 (cf. Lamb 1988). Caution is required. Regional wind direction and speed can be
modified at some sites by local topography, the geometry of forest and other obstructions, or
masked by the influence of local sea or lake breezes and valley winds. But marked local effects
can usually be detected and allowed for if the network is reasonably dense and given some
knowledge of the terrain.

As a first step in the map analysis, all the direct and proxy-weather information are plotted on
daily base maps. Using transparent overlays, the data for morning and evening hours are
transposed to separate charts. Each chart is then analyzed over the base map, which provides the
necessary background information as well as intermediate history, and in conjunction with
previous maps. Areas of cloud and precipitation are shaded in, the temperature and wind fields
studied, the zones of maximum gradient, wind shear and the pressure tendencies noted. Frontal
zones are tentatively indicated. Then an attempt is made to sketch in the pressure pattern, bearing
in mind the wind speed, the nature of the surface and the most likely direction and speed of

168

movement of the frontal systems. As an aid to analysis, where Registers exist, daily temperature
curves have been drawn up by month, with winds, cloud and precipitation added, to give a visual
display of the sequence of weather. All available pressure traces have also been plotted.

Here, the question of data calibration is partly resolved by the space and time smoothing implicit
in the map analysis. With respect to temperature, it is the relative differences and changes that
are important, and any significant errors would be expected to stand out. The results of the earlier
calibration study of the HBC data suggested that observing practices at this time were consistent
throughout the network; also that temperature readings were most reliable in the morning and
evening and near freezing. The wind-force scale in use on the Bay was quite similar to the
Beaufort Scale, and the latter has been used to convert all indicators to approximate speeds. Wind
direction was generally assumed to be with respect to true north. The barometer readings have
been reduced where possible to sea level.

Concerning the synchronization of data across Canada, the Hudson Bay region is considered as
the reference time zone; the western and eastern extremities of the map are then within about +
two hours, which can be born in mind. The timing and frequency of the observations are seminal
to the analysis. Problems here can often be overcome when maps are sketched in for both
morning and evening. In drawing up early weather maps, "historical continuity" becomes a prime
tool in grappling with the many difficulties, including that of sparse data. There is also the
advantage of knowing (if only in part) the future as well as the past.

Notes on Climate in Canada, 1809-20

What can be gleaned from these studies about climate in Canada from 1809 to 1820? Are there
any clues following the major volcanic eruption of Mount Tambora in April 1815 as to the
possible influence of such massive atmospheric loading on regional climate? To what extent was
the atmospheric circulation over east/central Canada anomalous at this time? The following notes
illustrate some of the kinds of climatic information that can be reconstructed using the Hudson's
Bay Company archives as an historical database.

Temperature

To set the climatic events of the period 1809-20 in context for the east coast of Hudson/James
Bay, Figure 3 shows the reconstructed series of May-October mean temperature anomalies
through the nineteenth century together with the modern twentieth century record. The reference
period is 1941-70. During the first decade of the last century, mean temperatures were similar
to those today, but then the seasons quite rapidly became cooler, tumbling in a spectacular fall
in 1816 and 1817 to values far below the modern records. Seasons well below the 1941-70
energy level remained characteristic of the first half-century with a cluster of cold years from
1835 to 1840; decadal averages for 1811-20 and 1831-40 were about 1.5°C below. A certain
mid-century amelioration was followed by some very cold weather in the 1860s, before a
remarkably sudden change occurred in the early 1870s to a new higher energy level akin both
to that at the beginning of the nineteenth century and to that today. Although not as warm as the
1870s, the last two decades sustained these milder conditions, and variability remained well
within that for 1941-70. The upward trend from 181 1-40 through 1870-1900 shows clearly in the
overlapping 30-year means. This sequence of events has no analogue in this century. The volcanic
eruptions of Tambora and Coseguina were followed by extreme cold, but the decline had set in
beforehand. For Krakatau and Agung, any similar signal (if that is what it is) is weaker, and for
St. Helens and El Chich6n, absent.

169

-0 08) (-0 03) (-0 04)

Tr (+0 1D (-0 10) (-0 07)

M (+0 3) (-16) (-11)8 (-15)" (-08) (-06)3 (-0 9)* (+0 1)* (-0.7)° (-0.2)

1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900
°C i I i I i I i i i I i I i I i I i I i °c

-0 06) (+013) (-0.01) (+0 07)

+ 3-|
+ 2
+ 1
0

-1 -

-2-

-3

-4-

-5-

-6-

1900

°C I

+3-

+ 2H

+ 1

-1 -

-2-
-3

TAMBORA

1910

LLI

[P [PIP

COSEQUINA

-MODERN INSTRUMENTAL RECORD

1920 1930 1940 1950 1960

KRAKATAU

1970

1980

1990

agungI l

r+3
2
-+-1
0
1

2
--3
- -4
--5
--6

2000

i °c

+ 3

-+2
- + 1
0
-1

--2
--3

(-+0 2)
(-0.3)

FORT GEORGE GREAT WHALE

(-0 1)

MT ST EL CHICHON
HELENS

1941-70 REFERENCE PERIOD

Figure 3: East coast, Hudson/James Bay. Reconstructed mean temperature anomalies (°C) summer
seasons (May-October) 1800 to 1900. M, decadal means; superscripts, number of years. Tr,
maximum tree-ring density, decadal mean anomaly (Parker et al. 1981). Anomalies with
reference to the 1941-70 period.

Within the May-October season, the most striking feature of the nineteenth century is the coldness
of spring (May /June) and autumn (September/October), and the effective shortening of the active
season. The curve for the spring months closely parallels that for the season as a whole, but in
autumn, cold seasons persist through the last two decades, although not below the modern record
low of 1974. In complete contrast, midsummers (July/August) were in general not so very
different from those today, although 1816 and 1817 with anomalies of about -5.5°C and -4.0°C
were in a class by themselves; cold summers also occurred in 1836 (-2.5°C) and 1871 (-3.0°C).
By far the coldest summer on modern record was 1965 (-3.5 °C), two years after Agung.

Focusing on the second decade of last century, with its remarkably sudden lowering of
temperature, Table 1 gives the absolute mean temperatures for 1808-20 (estimated for Great
Whale from the reconstructed regional anomalies) together with the 1941-70 normal values. The
first strike in the change to a colder mode was winter 1808-09, heralded by a very cold autumn
in 1808. Gladman, Master at Eastmain (a keen observer and a veteran of these shores), found it
the coldest winter he had ever experienced, with temperatures frequently below -40°C and
scarcely any mild weather. The next event was the extremely early and very cold autumn of
1811, advent of the even colder longer winter of 181 1-12; the two spring months May /June 1812
probably averaged below freezing at Great Whale (Table 1). These were considered extraordinary

170

times. In 1811, the HBC annual supply ship from England did not arrive at Moose Factory until
25 September, which was so late in the season and unprecedented at that time that there had been
great alarm at Eastmain lest no ship should arrive. As Gladman wrote: "these circumstances are
altogether so new and unfortunate" (HBC.B59/b/30). When the ship sailed for England, it could
no longer leave the Bay for ice, and wintered in Strutton Sound off Eastmain. An HBC supply
ship had last wintered over (though storm damage) in 1715 (Cooke and Holland 1978). It was
to happen again, as a result of ice, in 1815, 1816, 1817, 1819.

The most persistent and intensely cold period, continuously below the reference normal, began
in autumn 1815 and lasted until late winter 1818. The degree of cold reached its nadir in
January/February 1818, to be followed by a remarkable flip to an early, warm spring and a
benign season. Following this, cold springs and autumns continued. For 1815-20, Figure 4 shows
the monthly temperature anomalies for May-October at Great Whale, Fort George and Eastmain -
adjusted values based on temperature observations in the Meteorological Registers (cf. Wilson
1983b). Looking more closely at these seasons in Table 1 and Figure 4, several features are
outstanding.

1. The degree of the anomalous cold in 1816 and 1817, with many of the months below the
modern record - Alder, Master at Great Whale, then at Fort George: "if summers I may call
them" (HBC.B77/e/la). At Great Whale in 1816, July appears to have been nearly 6°C below
normal, that is 2°C below the lowest on record 1965. The modern standard deviation is
1.2°C. At Fort George in 1817, the season as a whole was some 5°C below normal. As a
season 1817 was more severe than 1816, but the greater severity was in spring and autumn
rather than midsummer.

2. From autumn 1815 through April 1818, the greater part of this coast experienced arctic
conditions following Koppen's definition. During summers 1816 and 1817, the arctic/subarctic
boundary lay close to Eastmain, some 3.5° latitude south of the present average position near
Richmond Gulf. The closest modern analogue is probably the 1965 season.

3. In 1815 and 1816, the rise of the daily mean temperature through 0°C, the start of the active
season, was two to three weeks later than the 1941-70 normal, akin to 1972. In 1817, it was
some four to five weeks late and the return through 0°C three weeks early in autumn, hence
the period above 0°C was some two months shorter. This was reflected in the break-up and
freeze-up of river ice. It is almost certain that seasonal ice remained in the ground in many
areas throughout the 1816 and 1817 seasons. It was also reflected in the snowcover, and type
of precipitation. The implications with respect to plant growth can be clearly seen in Table 1.

Precipitation

Again, to set the 1809-20 period in context, Figure 5 shows the series of May-to-October wetness
indices for the east coast of Hudson/James Bay through the nineteenth century, and the modern
record expressed in the same form. In general, the first half of last century was not only colder
but wetter than today, while the second half became warmer and drier. By far the wettest decade
was 1811-20, with a run of wet seasons from 1814 to 1820. The three wettest seasons of the
century 1816, 1817, 1820, at least matched the record in 1944. The three driest seasons were
1809 (which probably equalled the record 1920 season), 1807 and 1878. The rapid change from
the warmer/drier mode of the first decade to the very cold/wet regime of the second is
remarkable.

171

<u
a

>
£

a
I"

(/J

1820

—

iri

>/->

OnI

1^)1

NOI

Ttl

©|

i 1

NO'I

i 1

Ttl

noI

1816

qi
>ri|

On|

Ttl

ml
inl

Tt|

d
■ i

CTs|

c-i

r-'

'—'

ONI

d
• i

iri

oo'

Ttl

tri

1809

iri

°*l

1941-
1970

10.5

10.3

PERIOD

May to
October

May/

June

July/

August

Sept./
October

June/
August

August

u
c

•o

•c
8.

3
C
3

172

whale river big river

o|m|J|J|A|S|OM|J|J|A|S|oIm|J|J|A|S|o|mij|J|A|S|o|m|J|J|A|S|o|mijij|A|S|o|mij

1815

1818

M M

1820

[Register mis:

below modern
record

EASTMAIN

2 r

Figure 4: Whale River (Great Whale), Big River (Fort George), Eastmain, 1814-21: mean daily
temperature (adjusted values) expressed as differences from the 1941-70 normals (WR, BR) or
1960-72 averages (EM). The shading and asterisks indicate where the historical mean was
below the extreme monthly mean on modem record. M, data missing. These monthly
anomalies, together with the absolute values, are tabulated in the Appendix.

(-3) (+9) (+2)9 (+l)9 (+3) (-316 (-3)9 (-7)' (-3)9 (-7)
1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900
I , I I I i I i I i I i I i I i I i I . I

21 2

ffl

1900

1910

r— MODERN INSTRUMENTAL RECORD

1920 1930 1940 1950 1960 1970 1980 1990

I i I i I i I i I i I i 1 i I

2000
_J

(+4)7 o (+1>7 (+3)

* 1 FG,
VW

FORT
GEORGE

GREAT
WHALE

I s

tffl

ffl

fiP ffl B

vw
w

0
D

Figure 5: East coast, Hudson/James Bay. Wetness indices, summer seasons (May-October) 1800 to 1900.

Index from +2, very wet to -2, very dry. In parenthesis, decadal sums of the index;
superscripts, number of years. Asterisks indicate borderline cases, wet or dry; 1,2,3 wettest or
driest season. Reference periods: Great Whale, 1926-76; Fort George, 1916-69; Eastmain,
1960-76.

173

Looking at the detail in the Meteorological Registers and Post Journals for the seasons 1815-20,
the effect of such low temperatures on precipitation is evident. Figure 6 illustrates the greater
number of days with snowfall from May through October, and the shortening of the snow-free
season, contrasted with the modern period. The effect is especially noticeable on James Bay,
suggesting southward extension of the autumn snowbelt. Of particular interest is the summer
snowfall in 1816 at Great Whale; not only was there more snow in July than today, but even
more fell in August - a month that has no modern record of snow having fallen. Summer 1816
provides a marginal case for a residual snowcover on the east side of the Bay. In summer 1817,
snow conditions at Great Whale were most likely even more extreme. At Fort George, snowfall
was extraordinarily frequent and often heavy in May and June 1817, but no snow fell in July and
August, and the heavy rains of August probably washed away any snow remaining at the coast.

Thus the May-October seasons of 1816 and 1817 were such that had these conditions persisted,
they might have resulted in the formation of permanent snowfields in parts of New
Quebec/Labrador. From the impact of these seasons recorded in the Post Journals,
Correspondence and Annual Reports, it was indeed possible to see the southward expansion of
snow and ice forcing back the northern margins of habitation along this coast. The association
of volcanic activity and incipient glaciation is an old idea in the literature of climatic change.

GREAT WHALE (1941-70)

12 r-

M8 12 3 6 21

S 0

o nr
j i a I s '

FORT GEORGE (1941-70)

s — I* *p

S n 0 0 ° P

i u iMl J I J I A I S I O

EASTMAIN (1951-80)

S 0

m! J I J

WHALE RIVER

1814

►| BR

mIj! .

sIoImIjIjIaIsIoImIjIjIaIsIoImIjIjIaIs'oImIjIjIaIsIoImIjuIaIsIoImIjIjIaIsIoImIjIj

O 12

o
<r

lu 8
m
5
D

Z 4

EASTMAIN
1815

J I A I S I O

s lo MU I J I a I s I o m I j I j I a i s I o m I j I J I a I S I O Ml J I J I A I s io Ml J I j I a mI jIjIa'sio mIj'j

Figure 6: Whale River (Great Whale), Big River (Fort George), Eastmain, 1814-21: number of days with
snowfall, together with modern reference values for Great Whale, Fort George and Eastmain.
(The climatological day beginning at 8 a.m. at Whale River and Big River, 6 a.m. at Eastmain.)
M, data missing; J, Journal entries, no Register; *, less than 1 day.

174

Regional Climate and Tambora

Circumstantial evidence for the east coast of Hudson/James Bay from autumn 1815 to later winter
1818 suggests a possible case for regional climatic cooling through the intervention of Tambora.
That the material from the equatorial eruption in April 1815 should have entered the polar
stratosphere by autumn of that year, with a residence time of more than one year, is in keeping
both with the structure and behaviour of the atmospheric circulation and with studies of
radioactive fallout in the 1950s and early 1960s. Moreover, empirical studies and certain
theoretical considerations suggest that any resultant lowering of air temperature near the surface
in higher latitudes might be expected to be most apparent in the warm season, and of greater
magnitude than in lower latitudes. But in the case of Tambora (and of Coseguina), this appears
to be at most a short-term feature superimposed on longer-scale climatic changes. The onset of
cooling in this region occurred before the major event of Tambora (cf. Figure 3), with the return
to a warmer mode some 50 years later. While a number of smaller eruptions did take place in
the years preceding Tambora, there is also the concurrent event of the double cycle of abnormally
low sunspot number (the lowest since the Maunder Minimum), which spanned the first two
decades of the nineteenth century (Eddy 1976, p. 1 191; this volume). A further consideration is
the sudden swing to near-record warmth in April/May 1818 following hard upon the coldest
weather recorded; this almost suggests an "over-compensation" in redressing the balance. Had
there been a significant scavenging of the aerosol by the end of the very wet 1817 season?

For a different perspective on the 1816-17 seasons, it is useful to consider the regional climatic
controls and energy exchanges - in so far as clues are offered in the HBC archives - without
directly invoking Tambora.

Energy Exchanges: Advection

The thermostatic effect of the Bay on summer temperatures is a critical, if complex, influence
along this windward coast, where background air temperature level appears to be closely related
to the temperature of the Bay surface. Cold seasons do tend to have a higher proportion of Bay
winds, and this was the case in 1816 and 1817.

The arctic summers of 1816 and 1817 were marked by heavy ice and late break-up and melt in
the eastern and southern parts of the Bay. In both years, the last heavy ice was compacted in
southeastern Hudson Bay, extending into northern James Bay - a pattern similar to modern
maximum ice/water limits for mid-August to mid-September. In 1817, the timing was perhaps
a week later. But in 1816, these stages occurred some four weeks later between mid-September
and mid-October. In mid-September 1816, the ice in James Bay was akin to the normal for mid-
July. Given the cold autumn of 1816, the season provides a marginal case for the carry-over of
ice from one season to the next - a situation exceptional for the period itself. In contrast, the
clearing of the ice for the 1818-20 seasons was relatively early in this part of the Bay.

The windroses for Great Whale, June to August 1816, are shown in Figure 7, together with those
from the modern record. North winds are normally frequent here in May and early June,
associated with a series of anticyclones which cross from the Arctic into southeastern or eastern
Canada; this is then superseded by prevailing upper westerly flow with travelling depressions,
which is characteristic of summer. Figure 7 shows how the spring pattern continues through July
in 1816 with an unusually high frequency of north and west winds. In August the pattern
changes, but winds are overwhelmingly from the west. These months are dominated, then, by
the advection of cold air either from the Arctic or from passage over the Bay ice, and by the

175

50% 40% 30% 20%

20%

10%

20%
I—

No calms
reported

ure 7: Windroses for Whale River (Great Whale) 1816, and Great Whale 1942-54, 1967-76. (The
recent river bank sites are comparable with that of 1816.)

GREAT WHALE

MEAN (1953-72)

WARM MONTHS

M J J A S O

COLD MONTHS

I M J J A S O

WHALE RIVER

M I J I J I A I S'O

-WR— »-| BR
1816

BIG RIVER

No Register
J I A I S I O ! Ml J

S O

EASTMAIN
1815

I I A I S I O

40 O

M J J A S O

MJJ ASOMJ

ure 8: Whale River (Great Whale), Big River (Fort George), Eastmain, 1814-21: relative frequency
of "clear" hours, together with reference values for average, warm and cold months at Great
Whale. Morning, noon and evening hours. (A "clear" hour is defined by zero to five-tenths
cloud cover.)

176

almost total absence of the warmer southerly or land components. In 1816, as today, snow in
spring and early summer was brought by northerly and westerly winds. The persistence of
onshore winds in turn served to pack the ice in along this coast all summer, and further depress
coastal temperatures. At Fort George and Eastmain in 1817, the prevailing onshore winds from
June through August indicated the frequent passage of depressions. This suggests that the
exceptional cold of these two seasons was associated with different circulation systems.

Radiative Energy Exchanges

Today, low cloud is dominant in this region during the average summer season, and in
exceptionally cold summer months even more pronounced (Figure 8). In sharp contrast, a striking
feature of the seasons 1815-20 in general is the greater frequency of clear weather (zero to five-
tenths cloud cover) from May to August. In 1816 at Great Whale, the frequency was double what
might be expected today in very cold months, and particularly noticeable in July. Respective
listings of clear hours against wind direction and damp, cloudy hours with Bay winds suggest,
when compared with modern analogues, that the clear weather in 1816 was the result of: (1) the
prolonged influence of arctic airmasses at this period; (2) the greater frequency, persistence and
intensity of spring anticyclones over Hudson Bay, probably extending through July; and (3) the
late break-up and unusual persistence of heavy ice in the Bay through the summer.

Still leaving aside Tambora, the clearer skies in 1816 and 1817 compared with very cold summer
months today imply a larger receipt of incoming solar radiation at the surface, although the full
potential may have been reduced, particularly in the region, as a consequence of the unusually
"quiet" sun1. Considering the short-wave radiation balance, any increase in incoming radiation
at Great Whale in 1816 could have been more than offset by increased losses resulting from the
exceptional clarity of the air and through reflection from late snowcover and ice, enhanced into
July and from the third week in August by fresh snowfall. In the case of the long-wave radiation
balance, the dryness and clarity of the air and low sky temperatures would have encouraged loss
from any more favoured sites or surfaces, while the net radiation through the summer would have
been used primarily in melting snow and ice, and thawing and drying out the soil. These
conditions together with the low-level advection of cold air could go some way to account for the
very low air temperature at screen level at Great Whale in summer 1816.

Reintroducing Tambora, measurements of solar radiation following the eruption of El Chichtfn
in 1982 suggested a reduction in the total short-wave radiation reaching the surface as a result
of the stratospheric loading of dust and sulphur; while there is satellite evidence of enhanced
infrared emission from the cloud, the effect of the volcanic material on the long-wave balance
at the surface is not known.

To speculate, the evidence so far suggests that the cause of the extreme cold along this coast at
this time was most likely multiple: a combination of unusual external factors converging on the
years 1816 and 1817 (the general decrease in solar power, coupled perhaps with a high frequency
of heavy volcanic aerosol in this particular region of the stratosphere), whose climatic effects in
"summer" were magnified along this subarctic/arctic margin through the massive presence of

Given the apparent connection between auroral/geomagnetic activity and sunspots, it is worth noting that compared
with the high frequency of auroral activity observed in recent years, the only reports of auroral sightings in the HBC
Journals for this coast during the nineteenth century were in 1878, 1879 and 1880 (cf. Figure 3).

177

unusually late ice1 and snow, and the complexity of the ensuing surface - atmosphere
interactions.

Atmospheric Circulation, June-July 1816 - Preliminary Remarks

The maps from the sequence 1 June to 13 July 1816 are first approximations to illustrate the work
in progress (e.g., Figure 9). The results of the pilot study for 1-17 June (Wilson 1985c, 1985d)
had indicated that useful daily schematic flow patterns for east/central Canada can be drawn for
this early period, with the HBC archives providing the core database, supplemented where
possible by other historical weather sources. Two sample maps are reproduced here (Figures 9a,
b); additional information obtained more recently for the east coast of the United States and
Canada (cf. Figure 1) is now serving both as a check on the original analysis and to refine the
patterns. Although the HBC data are less complete in July, the coverage is still adequate when
the weather patterns are well-articulated, which was generally the case in the first two weeks
studied to date. At this early stage of the study, one or two preliminary remarks can be made
concerning the atmospheric circulation during the first half of summer 1816.

During much of this period, flow over east/central Canada and the northeastern United States was
predominantly meridional, interspersed by short periods of more zonal flow (notably the first
week in July) with rapidly moving depressions, and brief northward extensions of the Subtropical
High. The synoptic situation for the period 5-10 June points to blocking in the vicinity of Hudson
Bay; from 6 July until the end of the present analysis on 13 July, there is some evidence,
provided by the approaching HBC ships, of a blocking high east of Greenland.

The two exceptionally cold events over eastern Canada and the United States (5-10 June,
6-1 1 July) were apparently associated with these periods of blocking. In each case, a depression
passed across the Great Lakes/northern Ontario (cf. Figures 9a, 10), lost speed abruptly over
Quebec and developed into a large system, gradually drifting eastward. Behind the depression,
high pressure extended from the Arctic down over Hudson Bay, and very cold air was pulled
unusually far south in the rear of the storm. In June, the bitter northwest winds brought frost and
snow to the St. Lawrence Valley and New England (Baron, this volume). In July, the clear dry
air brought very low temperatures, especially at night, at least as far south as Philadelphia, where
it was as cool as late September - and mornings and evenings uncomfortably so2.

The intensity and size of some of the systems, as well as the highly variable and contrasting
extremes of temperature experienced throughout the region, bear witness to the vigour of the
north-south energy exchanges, and suggest a much stronger mid-latitude temperature gradient
than is usual today at this season. The storm tracks in early July were unseasonably far south.
A strong temperature gradient was present at times to the south of Hudson/James Bay between
the forested/spaghnum Shield country north of the Great Lakes, which on occasion became
extremely warm, and the unusually complete and compacted ice-covered surface of the Bay; here,
lows seemed to regenerate or develop.

The possibility of submarine seismic activity in the Arctic as a source of kinetic and heat energy, easing the breaking
up and outflow of previously compacted arctic ice, has still to be ruled out (cf. Wilson and MacFarlane 1986).
Deborah Norris Logan's diary, Historical Society of Pennsylvania.

178

Figure 9a: Surface weather map, 5 June 1816, morning. Temperatures in degrees Fahrenheit; in
parenthesis, mid-day values. Winds, short barb five knots, long barb 10 knots; broken arrow,
one observation a day, time unknown; asterisk, speed unknown. Pressure in millibars; recent
work has suggested that an adjustment of about +9 mb is required to reduce the station
pressure at Quebec City to sea level. (Reproduced with kind permission from Weather 10,
p. 137.)

Figure 9b: Surface weather map, 10 June 1816, morning. For legend, see Figure 9a. (Reproduced with
kind permission from Weather 10, p. 137.)

179

Figure 10: Trajectories of the surface high- and low-pressure systems, 3-13 June 1816. (Reproduced with
kind permission from Weather 10, p. 138.)

It appears that the stalled situation of 5-9 June was associated with the intensification of an
anticyclone over the Bay ice. On 9 June, the high began to move out to the southeast (cf. Figures
9b and 10) and was centred off the coast of New England by the evening of 11 June. Although
less persistent, this anticyclonic pattern of flow was repeated in the third week of June - the
trajectory now more southerly, over Lake Erie. The analysis is not yet complete, but a rather
similar situation may also have occurred at the time of the stalled circulation in July. Today, the
Hudson Bay High is characteristic of spring, (March-May) when anticyclones are more frequent
over eastern Canada. Although some of these spring anticyclones remain cold-surface features,
Johnson (1948) found that typically the Hudson Bay High is a deep system associated with a
warm ridge aloft in the vicinity of 90°W. This upper ridge is, in turn, related to a trough down
over western Canada and the United States near 1 12°W, while the trough off the east coast lies
near its normal spring position (about 55 °W over northern Newfoundland curving gently
south westwards). Such situations, which can persist up to five days or more, are preceded by
blocking over the North Atlantic and western Europe (see also Treidl et al. 1981). In addition,
Johnson noted that the trajectories of the highs shifted during the course of the season from
southeastward over New England in March and April to a more southerly route in May.

180

Thus the map evidence strongly suggests that the atmospheric circulation over eastern Canada and
the northeastern United States in June and the first half of July was abnormal in its timing rather
than in kind - that spring had been delayed or protracted by as much as six weeks. This is in
keeping with the evidence of the regional climate on Hudson/James Bay, where the normally brief
summer was essentially obliterated. It points yet again to the importance, climatically, of the
extraordinary ice and snow conditions over the Bay and northeastern Canada in summer 1816,
and more generally to the key role of Hudson Bay at this season in the climate of this part of
North America.

Concluding Remarks

For the east coast of Hudson/James Bay during 1809-20, the evidence associating the extreme
cold with the eruption of Tambora remains circumstantial and intriguing. It is tempting to see the
period of unrelieved record cold from October 1815 to March 1818 as influenced by high levels
of stratospheric dust and acid - a relatively short-term feature exacerbating a longer-term
(60-year) lowering of temperature, which had already begun. There can be no doubt that in 1816
and 1817 the combination of exceptionally heavy ice lingering through the summer in this part
of the Bay and the high frequency of onshore north and west winds was a major factor in
reducing coastal air temperature through the "warm" season.

From a different perspective, the daily surface weather maps reconstructed for east/central
Canada, from 1 June to mid-July, indicate circulation patterns and sequences normally related to
ice and snowcover over the Bay and northeastern Canada, and suggest that spring was running
some six weeks late in 1816.

The riches offered by the HBC archives in terms of Canada's historical climate have scarcely
been tapped. A major inhibiting factor is the labour-intensive, time-consuming nature of the
work. The great challenge is to reduce the time and labour required without sacrificing
information, quality control or physical reality - truly a worthy challenge to modern computer
techniques.

Acknowledgements

Warmest thanks to Gordon McKay, Howard Ferguson and Mai Berry of the Canadian Climate
Centre for long-term support since 1977 through a series of contracts, which have made the bulk
of this work possible; also to my Scientific Officers Bruce Findlay and Joan Masterton, to Valerie
Moore for processing all the manuscripts, to draftsman Brian Taylor (cf. Figures 4,6,7,8), and
to the many others at the Atmospheric Environment Service who have helped me over these
years.

I am grateful to Dick Harington of the National Museum of Natural Sciences (now Canadian
Museum of Nature) for bringing me into the Museum's Climate Change in Canada Project, and
for his encouragement particularly with the synoptic-mapping study, for which the Museum
provided seed moneys. I appreciate too the financial aid and the other help that he has provided,
with Gail Rice, in publishing my papers in Syllogeus, and the work of Edward Hearn (Ottawa
University) who has drafted nearly all my figures for Syllogeus, most under contract to the
National Museum.

181

I also thank the Hudson's Bay Company for permission to use the Company archives, HBC
archivists Joan Craig and Shirlee Smith and their staff, and Alan Cooke, who in 1965 as a
colleague at the Centre d'Studes nordiques, University Laval introduced me to the climatic
material contained in these archives, thereby opening up a new world.

References

Ashmore, S.E. 1952. Records of snowfall in Britain. Quarterly Journal of the Royal
Meteorological Society 78:629-632.

Cooke, A. and C. Holland. 1978. 772*? Exploration of Northern Canada, 500 to 1920, A
Chronology. The Arctic History Press, Toronto. 549 pp.

Eddy, J. A. 1976. The Maunder Minimum. Science 192:1189-1202.

Geiger, R. 1965. The Climate Near the Ground. Harvard University Press, Cambridge,
Massachusetts. 611 pp.

Johnson, C.B. 1948. Anticyclogenesis in eastern Canada during spring. Bulletin of the American
Meteorological Society 29:47-55.

Kington, J. 1988. The Weather of the 1780s over Europe. Cambridge University Press. 166 pp.

Lamb, H.H. 1988. The weather of 1588 and the Spanish Armada. Weather 43:386-395.

Landsberg, H. 1967. Physical Climatology. Third edition. Gray Printing Company. Dubois,
Pennsylvania. 446 pp.

Manley, G. 1978. Variations in the frequency of snowfall in east-central Scotland, 1708-1975.
Meteorological Magazine 107:1-16.

Parker, M.L., L.A. Jozsa, S.G. Johnson and P. A. Bramhall. 1981. Dendrochronological studies
on the coasts of James Bay and Hudson Bay. In: Climatic Change in Canada 2. C.R.
Harington (ed.). Syllogeus 33:129-188.

Pfister, C. 1980. The Little Ice Age: thermal and wetness indices for central Europe. Journal of
Interdisciplinary History 10:665-696.

Treidl, R.A., E.C. Birch, and P. Sajecki. 1981. Blocking action in the northern hemisphere: a
climatological study. Atmosphere-Ocean 19:1-23.

Wilson, C. 1982. The summer season along the east coast of Hudson Bay during the nineteenth
century. Part I. General introduction; climatic controls; calibration of the instrumental
temperature data, 1814 to 1821. Canadian Climate Centre Report No. 82-4:1-223.

. 1983a. Part II. The Little Ice Age on eastern Hudson Bay; summers at Great Whale, Fort

George, Eastmain, 1814-1821. Canadian Climate Centre Report No. 83-9:1-145.

182

. 1983b. Some aspects of the calibration of early Canadian temperature records in the

Hudson's Bay Company Archives: a case study for the summer season, eastern
Hudson/James Bay, 1814 to 1821. In: Climatic Change in Canada 3. C.R. Harington
(ed.). Syllogeus 49:144-202.

.1985a. The summer season along the east coast of Hudson Bay during the nineteenth

century. Part III. Summer thermal and wetness indices. A. Methodology. Canadian
Climate Centre Report No. 85-3:1-38.

. 1985b. The Little Ice Age on eastern Hudson/James Bay: the summer weather and climate

at Great Whale, Fort George and Eastmain, 1814-1821, as derived from the Hudson's
Bay Company Records. In: Climatic Change in Canada 5. C.R. Harington (ed.).
Syllogeus 55:147-190.

. 1985c. Daily weather maps for Canada, summer 1816 to 1818 - a pilot study. In:

Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus 55:191-218.

. 1985d. Daily weather maps for Canada, summer 1816 to 1818. Weather 40: 134-140.

. 1988. The summer season along the east coast of Hudson Bay during the nineteenth

century. Part III. Summer thermal and wetness indices. B. The indices 1800 to 1900.
Canadian Climate Centre Report No. 88-3: 1-42.

Wilson, C. and M.A. MacFarlane. 1986. The break-up of Arctic pack ice in 1816 and 1817.
Weather 41:30-31.

183

i "8
I" t

C o

p»

c i

NO
00

00
00

ON
00

o
r i

M in ^ o> n
O O rn O O

NO ( — <

NO ON

>/-> no

no r- on oo r)
© ■>* —

On tj- rj ~ vn Tt
« ri Tt oo no 6

r- O cn on cn >n
O — ^ Tt vri — o

'1 i .1

S E o
2 £ B

u
o

S £ £

vn no
cn ri

^ oo in 't

on O ri

r» </->
ci — ri
+ ' '

ri
+

r»

rl on r- r~ no m m
cn no O ri O ri

cn 00 r) On On NO ©

wi cn od o — cn
'''+'''

TfiriNor^oooNO'-

00 00 00 00 00 00 00 00

« -<t m O no

ri ri no ri ri

cn — i

r- cn

o On VT) no 00

o 1- d -

On NO r-; rn no
•<t cn -»t o ri — i

[ l T T l +

no oo Tt r» o
— ri no cn —i ri

l i T + l ,

c © —

.2 o
CQ W

vn no r~
oo oo oo

— ■ r)
ri «-<

no r~
6 6

c o

s p

•o

3
o
c
c

Oil
u

8
a
t:

>
m

"t3

u
a:

2 =

u .a
oa -o

H <

184

Climatic Change, Droughts and Their Social Impact: Central Canada,
1811-20, a Classic Example

Dr. Timothy Ball1

Abstract

Changes in the climate of central Canada from 1760 to 1800 were marked by extreme fluctuations
as the region began to emerge from the nadir of the Little Ice Age. The harshness of climate,
particularly along the northern limit of trees, created severe ecological conditions. Evidence from
the historical and meteorological records, maintained primarily by the Hudson's Bay Company,
provides clear indications of the extremes and the impact climatic changes had on the socio-
economic infrastructure of the region. Between 1800 and 1810 the climate was relatively benign,
holding promise of better conditions and times in the nineteenth century. The promise was short-
lived as temperatures began to decline in 1811, a trend that was to continue through to 1818.
Most research on the period has stressed the temperatures, but detailed studies of the historical
documents show that the period from 1815 to 1819 was one of severe drought as well as cold.
The combination suggests that the mechanisms causing the drought were probably different than
those that created the hot droughts of the 1930s.

Droughts in the Canadian prairies are usually attributed to a northward extension of the Pacific
High (Subtropical). Droughts in the boreal forest region are usually associated with a southern
position of the Arctic High. Both regions indicate a 22-year cycle of droughts that seems to
coincide with sunspot cycles. With a northward shift of the mean summer position of the Arctic
(Polar) Front there is a hot drought on the prairies. With a southerly location of the Front there
is a cold drought. The 1815-17 period was a classic example of the latter, and is associated with
an extreme degree of meridionality in the zonal index.

Variations in precipitation in the early nineteenth century had a significant effect upon the wildlife
of the region. This resulted in a decrease in the food supply for Europeans and Indians, with the
concomitant social and health stresses. A decline in the fur-bearing animals created declines in
income that led to significant socio-economic adjustments.

Introduction

A great deal of attention has been paid to the exceptionally cold summer of 1816. A study of the
eventful summer has been documented by Hoyt (1958) with a description of weather, food
supplies, prices, and even population movements. Work focused upon the northeastern United
States and Canada. Attention spread to what Post (1977) called, "the last great subsistence crisis
of the western world". The most extensive and analytical study was the book Volcano Weather
(Stommel and Stommel 1983). This work included a chapter on conditions reported in Europe.
A brief reference at the end of the chapter suggests limits to the extent of the area influenced by
cold conditions, "It would appear, however, that the truly exceptional character of 1816 weather
was limited to a small portion of northeastern America, Canada and the extreme western parts
of Europe" (Stommel and Stommel 1983, p. 51).

Department of Geography, University of Winnipeg, 515 Portage Avenue, Winnipeg, Manitoba R3B 2E9, Canada.

185

The severe cold conditions have more recently heen detailed in the Hudson Bay region of
northern Canada (Catchpole 1985; Catchpole and Faurer 1985; Skinner 1985; and Wilson 1985).
Such work has produced not only valuable information on the extent and intensity of the cold
conditions in 1816, but it has also contributed to ideas concerning the possible causes of such
exceptional conditions.

The debate centres on whether the cold conditions were caused by: (1) the intensity of the dust
veil emitted by the equatorial eruption of Mount Tambora in April 1815; or (2) the influence of
variable sunspot activity; or (3) natural variation caused by some atmospheric, or
atmosphere/ocean phenomenon. This controversy remains unresolved. While each effect may
exert an influence on large-scale atmospheric circulation, by reducing the Earth's radiation
balance, they may also exert a simultaneous effect.

Documentary evidence for the summer of 1816 appears to agree on one aspect of the cold
summer - that a ridge of high pressure extended south over eastern North America and western
Europe bringing cold arctic air well south of its normal latitudes for the time of year. These
systems are relatively common in the fall, winter and early spring, but are unusual in the
summer. Their impact on the socio-economic conditions of that period were severe.

Here, I intend to show that the pattern of weather in 1816 can be generally defined from climatic
information in Hudson's Bay Company records. The pattern indicates that cold conditions did not
include the entire prairie region. Southern Alberta had normal conditions, while the north had
an exceptionally wet summer. Overall weather conditions began to deteriorate in 1809, and
continued to decline until 1816. That year, apart from being cold, was the first year of a severe
drought that lasted until 1819. Comparison of conditions with modern synoptic charts suggest that
this was a cold drought within the 22-year cycle of droughts experienced in the Great Plains.

The 22-year cycle of droughts correlating with sunspot activity has generally been established and
accepted (Herman and Goldberg 1978). Very little detailed analysis of the nature of each drought
period has been completed. It is generally accepted that droughts are coincident with hot weather,
and the 1930s are cited as the classic example. Undoubtedly hot dry weather is especially
damaging to modern agriculture, but lack of precipitation under any temperature regime is
serious. The cold drought from 1816 to 1819 was especially damaging, as journals and diaries
record.

These years of severe weather had a considerable impact upon wildlife, indigenous peoples and
Europeans. Later, I will suggest that it served as a catalyst for an already volatile situation: the
Seven Oaks massacre in 1817 at the Red River settlement.

The fur trade had been suffering from over-trapping and competition between the North West
Company and the Hudson's Bay Company. Tensions between Indians, Mens and fur traders were
somewhat overshadowed by the growing talk of permanent European settlement. The first group
to arrive, the Selkirk Settlers, came from Scotland in 1811. Generally, they were unwelcome
because they threatened the fur trade and the traditional ways of native people. Ironically, they
had left Scotland because of severe weather. Now they had moved into a land that was suffering
for the same reason. Diminished wildlife populations meant reduced food supply, with associated
hunger and disease. Residents already felt threatened and unsure, thus they saw the settlers as an
even greater threat. This situation heightened tensions and began a long period of conflict.

186

The Hudson's Bay Company Post Journals often provide a brief but daily summary of activities
and weather conditions. Some Journals are incomplete in this period because of feuds between
the Hudson's Bay and the North West companies, or absences from the Posts on expeditions
between Inland Posts and Bay Posts to exchange furs for supplies. Sometimes the mere struggle
for survival precluded maintenance of the records. Although the Journals are fragmented, the
entries provide a series of proxy data that give some indication of the activities and weather
conditions that occurred within these years.

Data Sources

Proxy data from the Journals include: (1) comments on garden-crop preparation, growth and
damage; (2) remarks of frost or ice formation, movement and decay; (3) descriptive terms for
winds or precipitation restricting outdoor activities or travel. Even limited comments of
phenological data for animal appearances or migrations serve as an indicator of weather
anomalies.

The Hudson's Bay Company Posts located in the west-central region of Canada are shown in
Figure 1. Brandon House was an important Post adjacent to the Assiniboine River. Peter Fidler,
the Factor in charge, provided much of the proxy data and insights into the weather through his
records. Carlton House, although shifted several times, was transferred in 1810 to a site near "a
crossing place" on the south bank of the North Saskatchewan River. The daily accounts written
for each Post between May and October 1810 to 1820, were analyzed for proxy data that might
be attributable to adverse weather.

Figure 1: Location of Hudson's Bay Company posts.

187

Gardening and crop production were a major part of the general way of life. The Company
encouraged each Post to obtain much of its sustenance from local sources in order to be as self-
sufficient as possible. Gardens were maintained, and hunting and fishing supplemented the diet
with a fresh supply of meat. Cutting firewood for the approaching winter also took up much of
the time (Ball 1987). Frequent canoe trips to Posts along Hudson Bay were embarked upon in
the short summer season. The observations recorded during these activities are evidence of the
severe cold experienced in the early to late summers of 1816 and 1817. Warmer weather returned
in 1818, but the drought continued to 1819.

Peter Fidler's daily records for Brandon House yield valuable information. In 1816, the ice must
have dispersed toward the end of April or early May as Indians were fencing in the Assiniboine
River on 21 May "...half of mile above the House to kill sturgeon" (HBCA, PAM B22/a/19).
Fidler mentioned that the presence of sturgeon usually gives an indication of when the ice goes
out "...they annually come up every spring in great numbers when the ice goes away and they
appear here about 10 to 12 days after it clears away..." (HBCA, PAM B22/a/19)'.

Gardening and crop preparation began on 3 May, and an indication of the spring runoff levels
and weather were observed by Fidler on 9 May as water levels were "...falling daily 1% inch -
cold weather and strong wind these two days" (HBCA, PAM B22/a/19). The remainder of the
crops were sown by 27 May, but it was not until 5 June that a severe cold spell occurred. "A
very sharp frost at night and killed all the Barley, Wheat, Oats and garden stuff above the ground
except lettuce and onions - the Oak leaves just coming out are as if they are singed by fire and
dead" (HBCA, PAM B22/a/19). With a severe frost early in June the growth of crops and natural
vegetation would certainly be curtailed, as this period is essential for their development to mature
plants.

An interruption in the Journal occurs after this period due to the battle between the Hudson's Bay
Company and the North West Company2. In the spring of 1817, Brandon House was subjected
to severe weather, as on 18 June "...thin snow fell 2 inches deep" (HBCA, PAM B22/a/20).
Visitors were late arriving at the Post due to the backward spring, and Fidler explained the cause
of late arrival; "...he was detained long by the ice in the Little Winnipeg" (HBCA, PAM
B22/a/20). Little Winnipeg refers to Lake Winnipegosis. The summer of 1817 was reported to
be backward due to the unseasonable cold, but drought also had a direct effect on agriculture.
"The crops exceedingly backwards - some potatoes only 4 inches above ground - whereas in other
seasons there were new ones bigger than Walnuts, the grass is also remarkably short and ground
dry - all the little runs of water now dry - so there is every reason to expect a bad crop on
account of the great want of rain - the season has been colder than usual" (HBCA, PAM
B22/a/20).

The summer continued to be dry, as the small saline lakes began to evaporate. The apparent
migrations of buffalo southward also give an indication of the dry summer and unusual cold.
Fidler recorded buffalo movements near the Post in the spring, and by 11 August they were
"...very numerous - even extending so low down as the Forks" (HBCA, PAM B22/a/20). The
cold weather continued. Fidler wrote in August of frosts that occurred on 17 and 23 July at
Brandon which killed all the potato tops. The autumn season seems to end on 23 October when

1 Hudson's Bay Company Archives, (HBCA) Provincial Archives of Manitoba (PAM), Journal Number.

2 The conflict between the two companies was resolved in 1821 when the smaller Hudson's Bay Company
incorporated the North West Company.

188

the Assiniboine River froze over. This occurrence is seen by Fidler as being "...very early in the
season, about 20 days sooner than usual - and it set in early last fall" (HBCA, PAM, B22/a/20).

The spring of 1818 apparently began without mention of adverse conditions, as Fidler reports on
26 May "...the ice drove by about 5 weeks ago ..." (HBCA, PAM, B22/a/21). Despite a break
in the daily reports for Brandon House, Fidler continued to write on his journey from Red River
to Martins Falls near Albany Factory on James Bay.

Returning to Brandon House, Fidler recorded the late summer conditions: "Water very low in
the river and a very dry season scarce a single shower of rain all summer, all the potatoes and
garden stuff quite burnt out as also Vh bushels of Barley sown there - when 3 inches high all
killed by the great drought - these 3 summer past remarkably little rain ... quite different from
what it used to be" (HBCA, PAM, B22/a/21).

Summers at Carlton House are also recorded in fragments due to continuing battles between
Hudson's Bay and North West companies, and canoe trips to other Posts. However, direct
information recorded in the Journals still provides an indication of summer conditions for
1816-18.

The spring of 1815 appeared to have a positive beginning: crops were planted as early as
29 April. However, conditions changed, and on 13 May John Pruden recorded the bleakness of
the weather; "...hard frosts every night retards vegetation very much, none of the seeds that have
been sown make their appearance above ground except the cabbage seed" (HBCA, PAM,
B27/a/4).

By June, weather continued poor, but there was a different problem. "Wind SW blowing fresh
part-clear and part cloudy weather, it has been remarkable dry wind weather all this month which
keeps the garden stuff very backward" (HBCA, PAM, B27/a/5,2d). Things had not improved a
month later: "The insects have eaten all our cabbage and turnips owing I suppose to the dry
season, ..." (HBCA, PAM, B27/a/5,4). This was the first indication of a drought that was to grip
the eastern half of the prairies for three years. The impact was to be quite severe.

The drought conditions are best summarized in Peter Fidler's General Report of the Red River
District for 1819.

The spring months have sometimes storms of wind and thunder even so early as
March within these last years the Climate seems to be greatly changed the
summers so backward with very little rain and even snow in winter much less than
usual and the ground parched up that all summer have entirely dried up, for these
several years loaded craft could ascend up as high as the Elbow or Carlton House
but these last 3 summers it was necessary to convey all the goods from the Forks
by land in Carts... (HBCA, PAM, B22/e/l,6).

We can discover the extent of the drought by noting which rivers are reported to be low. The
North Saskatchewan, Assiniboine, Red, Hayes, Nelson and Steel rivers all receive attention in
the journals. This means that the drought covered the drainage basins of all of these rivers, thus
encompassing a large part of central North America.

The degree of the drought can be determined by the impact that it had on the environment,
wildlife and subsequently the people. James Sutherland reports that water routes connecting the

189

Hayes and Nelson rivers were only made passable by the construction of dams (HBCA, PAM,
B154/e/l,2). Peter Fidler notes that:

...as the country wherever I have been and by the invariable information of the
different Tribes I have enquired at agree the country is becoming much drier than
formerly and numbers of small lakes become good firm land will be covered with
Timber of various kinds. ..(HBCA, PAM, B22/e/l,8d).

Fidler implies that he expects these conditions to persist in the future, although he does not
specify for how long.

The value of his comments lie in putting the individual events into a larger and longer climatic
framework. It is important to note that all seasons suffered from the lack of precipitation. "These
3 summers past remarkably little rain - as also very little snow in winter quite different from what
it used to be" (HBCA, PAM, B22/a/21,29d). We also know that conditions were good prior to
the drought; "...since 1812 there was always good crops of everything until 1816 when the dry
summers commenced..." (HBCA, PAM, B22/e/l,8).

Prairie droughts are usually accompanied by the appearance of insects, particularly grasshoppers,
that exacerbate the problems. Fidler makes some interesting comments when talking about the
grasshopper infestation. He notes that, "...They first made their appearance the third week of
August 1818 at 2 O'clock in the afternoon and came from the southwest" (HCBA, PAM,
B22/e/l,20). The direction is significant because it indicates wind direction during the period.
This is confirmed by John Pruden's observation at Carlton House that, "Wind SW blowing fresh
part-clear and part cloudy weather, it has been remarkable dry wind weather all this month..."
(HBCA, PAM, B27/a/5,2d). Then Fidler writes: "These insects {grass-hoppers) make their
appearances in great numbers about every 18 years..." (HBCA, PAM, B22/e/l,6d). This implies
cycles of infestation and possibly of climate.

Atmospheric Circulation

So far we have established that 181 1-20 had below normal temperatures, especially in the years
1816 and 1817. Between 1812 and 1816, conditions were cool but generally good for crops and
vegetables. In 1816 drought began in a large region including the central and eastern prairies.
The drought ended in 1820 as temperatures and precipitation patterns returned to long-term
normals.

How did the circulation pattern for these years differ from the long-term normal, and was the
drought typical of those that occur regularly on the prairies?

The climate of central Canada is generally determined by the position of the Arctic Front1. In
summer the mean position of the Front approximates the northern boreal forest limit. In winter
it curves south in a great arc toward the centre of the continent to an approximate mean position

There appears to be some confusion over the use of the term Arctic Front. Bryson and others have used the term
Arctic Front to describe the major division between Arctic and Temperate air in North America. The term Polar
Front is used by others, particularly in Europe, presumably to indicate that there is a similar front in the southern
hemisphere. I have used Arctic Front because the paper is examining conditions in North America.

190

of 40°N Latitude. This curve occurs because the Rocky Mountains act as a barrier, and create
a standing wave in the westerly flow of the general circulation.

Polar air north of the Arctic Front tends to be cold and dry, while subtropical air to the south
tends to be warm and dry. Generally, moisture is brought to the region by cyclonic storms that
move along the front; these usually occur in spring and autumn as the Front moves through the
region in its annual migration. Most summer precipitation is convectional as instability develops
in the warm subtropical air.

Dey (1973) analyzed synoptic conditions occurring during summer dry spells in the Canadian
Prairies. He showed that the most severe droughts occurred when the Pacific High (subtropical)
extended northward into the southern prairie region. Blocking occurs with the extreme meridional
pattern that is formed. This configuration is commonly called an 'omega block'. Low pressure
zones on the Pacific coast and in the region to the west of Lake Superior lie on each side of a
large high pressure region. On the weather map this creates a pattern similar to the Greek letter
omega - hence the name.

It is also possible to have dry conditions in summer if the Arctic Front extends southward over
the region. This would produce cool, dry conditions under a predominantly northerly flow. The
summers of 1815-17 are good examples, being marked by very cool dry conditions in the early
summer as the Arctic Front remains well south of its normal position. When the Front finally
retreats northward, the prairie sites experience the associated precipitation. For example, at
Carlton House in July 1815, the Journal reads: "Wind easterly cloudy weather had a heavy
shower of rain last night, the only one I may say since summer commenced..." (HBCA, PAM,
B27/a/5,3). In June 1817, Swan River experienced three days of continual rain under cyclonic
conditions as the Front migrated northward.

Further evidence to support this hypothesis is provided by the weather patterns at Fort
Chipewyan. This post is ideally located to determine the latitudinal and longitudinal shifts of the
Arctic Front in summer. Spring was late each year from 1815 to 1818 inclusive. The summers
were cool and short, as autumn came early. This was especially true in 1816 as the entry for
30 September indicates: "One of the several days that I have witnessed at this season of the year,
the ground covered with snow Vh inches deep, blowing very fresh and extremely cold" (HBCA,
PAM, B39/a/9,10d). 1818 saw the return to a longer summer, with the first snow falling on
13 October, and a comment that there was "...mild weather with wind from the south" on
22 October (HBCA, PAM, B39/a/14,8d).

Ile-a-la-Crosse, south of Fort Chipewyan, has a limited record, but it does indicate normal
conditions. For example, the earliest date of the water being clear of ice in the modern record
is 12 May. In 1816 the ice was gone by 18 May. It is reasonable to use this station as the western
limit of the outbreak of cold arctic air in the spring and early summer of that year.

In summary, it appears that 1815 saw the beginning of generally cooler conditions. The cold was
more notable in 1816 and 1817, especially for the spring and early summer in the eastern half
of the Prairies. The Jetstream and Arctic Front swung south so that the eastern half of the Prairies
was cold and dry, under arctic air. Conditions changed significantly in 1818. The Arctic Front
moved north, and an omega block system set in that appeared to dominate through 1819. Thus
the cold drought that existed in 1816 and 1817 was replaced by a warm drought in 1818 and
1819.

191

Figure 2: General reconstruction of the pressure patterns for North America and the North Atlantic for
July of 1816 (after Catchpole 1985; Lamb and Johnson 1966).

Figure 3: Surface weather map: morning, 5 June 1816 (after Wilson 1985).

192

Figure 4a: Synoptic weather pattern for drought conditions in Western Canada (after Dey 1973).

Figure 4b: Synoptic weather pattern for drought conditions in Western Canada (after Dey 1973).

193

This pattern is consistent with the general reconstruction suggested by Catchpole (1985)
(Figure 2). It also demonstrates that Wilson's (1985) diligently drawn synoptic maps for three
months of the summer of 1816 are valid. Wilson's map for 5 June 1816 (Figure 3) shows the
surface conditions with a southern expansion of the Arctic High. Synoptic conditions as
reconstructed by Dey (1973) would have been the general situation in 1818 and 1819 (Figure 4).

Climatic Impact

Regardless of the climatic mechanisms, there is no doubt that climatic conditions seriously
affected wildlife and people's ability to grow food. The lack of precipitation reduced the planted
crops, but it also affected the wild harvest of berries and other fruits. Lack of snow is devastating
to many wildlife species, especially with colder-than-usual temperatures.

Climate also affected people's ability to travel. Early ice, late ice, too little snow, as well as
shallow rivers and lakes all hampered movement for trade and hunting. Buffalo migration also
indicates unusual conditions. Normally these animals would move westward or southward, but
with changing snow patterns and dry conditions to the west they altered their behaviour. Fidler
writes in 1817: "There are plenty of Buffalo not 15 miles off and all last winter and this spring
they have been very numerous - extending even so low down as the Forks" (HBCA, PAM,
B22/a/20,9d). Again in 1818: "The Catfish during these summers have also been very
scarce... but fortunately vast numbers of Buffalo have kept pretty near all summer" (HBCA,
PAM, B22/a/21,2).

Life has always been a struggle in this part of the world. Food supply varies dramatically with
climate. It is truly a region of plenty or dearth. However, the 181 1-20 period was one of special
severity. The 1780s and 1790s had been periods of severe climatic conditions, with weather
oscillating from one extreme to another. The author has argued elsewhere that this period created
the pressure that forced the Company away from its complacent position on the Bay. A brief
respite from 1800 to 1810 was then shattered by severe cold and drought. Starvation and hardship
returned, as animals disappeared or changed their routines. Fur traders saw their industry
threatened. Indians saw the fur trade threatened and their traditional way of life dashed. Conflict
for the trade was at a peak as the Hudson's Bay and North West companies literally battled over
the spoils. The Selkirk Settlers entered the scene unaware of the tensions and problems. Since
the settlers posed a threat to all, it is not surprising that a series of confrontations occurred
culminating in the massacre at Seven Oaks in 1817, when 20 men were killed. The stress of the
impending social and cultural confrontation was underlain by exceptionally severe weather.
Starvation and malnutrition made rational behaviour less likely.

Conclusion

There appear to be two types of drought on the Great Plains of North America; hot droughts and
cold droughts. The former coincide with the omega blocking system that dominated the region
in 1988. In this system the Pacific High extends northward over southern Alberta, southern and
central Saskatchewan, and southern Manitoba. This creates very hot and dry conditions similar
to those seen in the 1930s and again in the 1980s. The latter occur when the Arctic Front dips
south across these same regions so that they are now dominated by clear, cool and equally dry
conditions.

194

The 1816-19 period was one in which cold drought predominated. Journals of the Hudson's Bay
Company provide much information about the extent and intensity of the conditions. They also
allow estimation of the impact that these conditions had upon the environment, and therefore upon
wildlife and people.

The heat and drought of the 1980s have led to current predictions of global warming and
impending doom as droughts increase in frequency and severity in North America. My brief
study suggests that this will not be the case. Perhaps the pattern of hot or cold droughts will
change. A more northerly location of the Arctic Front might result in less southerly incursions
of Arctic air.

References

Ball, T.F. 1987. Timber! Beaver 67(2):45-56.

Catchpole, A.J.W. 1985. Evidence from Hudson Bay Region of severe cold in the summer of
1816. In: Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus 55:121-146.

Catchpole, A.J.W. and M.-A. Faurer. 1985. Ships' Log-Books, sea ice and the cold summer of
1816 in Hudson Bay and its approaches. Arctic 38:(2) 12 1- 128.

Dey, B. 1973. Synoptic climatological aspects of summer dry spells in the Canadian Prairies.
Unpublished Ph.D. thesis. University of Saskatchewan. Saskatoon. 180 pp.

Herman, J.R. and R.A. Goldberg. 1978. Sun, Weather and Climate. Scientific and Technical
Information Office, NASA, Washington, D.C. Sp-426. 360 pp.

Hoyt, J.B. 1958. The cold summer of 1816. Annals of the American Association of Geographers
48:118-131.

Lamb, H.H. and A.I. Johnson. 1966. Secular variations of the atmospheric circulation since
1750. Geophysical Memoirs 110. H.M.S.O., London. 125 pp.

Post, J.D. 1977. The Last Great Subsistence Crisis in the Western World. Johns Hopkins
University Press, Baltimore. 240 pp.

Skinner, W.R. 1985. The effects of major volcanic eruptions on Canadian climate. In: Climatic
Change in Canada 5. C.R. Harington (ed.). Syllogeus 55:75-106.

Stommel, H. and E. Stommel. 1983. Volcano Weather. Seven Seas Press Inc., Newport, Rhode
Island. 177 pp.

Wilson, C.V. 1985. Daily weather maps for Canada, summers 1816 to 1818 - a pilot study. In:
Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus 55:191-218.

195

The Year without a Summer: Its Impact on the Fur Trade and History
of Western Canada

Timothy F. Ball1-2
Abstract

Edward Umfreville referred to the Hudson's Bay Company as "being asleep by the frozen sea".
He was talking about the fact that the Company had established its trading posts along the shores
of Hudson Bay and made no attempt to build permanent posts inland. Arthur Dobbes used this
as evidence in his charge of monopoly against the Company. He claimed that the Company was
deliberately protecting and hiding the potential of the interior of North America to ensure the
dominance of the Hudson's Bay Company.

At the end of the eighteenth century the Company established its first inland post at Cumberland
House on the Saskatchewan River. It has always been argued that the sole reason for this move
was to counteract the expansionism of the North West Company. Increasing evidence suggests
that climatic change brought about a dramatic decline in the ecology of the northern region and
this was a major cause of the move inland. By 1810 the expansion created increasing conflict
between the two companies. In 1812 a third component, the Selkirk Settlers, arrived and the
turmoil continued to build.

Severe weather affected all three groups through their dependence upon the land for sustenance
and economic profit. Two events, the Seven Oaks Massacre and the amalgamation of the
Hudson's Bay and North West companies, followed the period of most severe weather in 1816-
17. There is little doubt that the summer of 1816 was one of the worst in the historic record. It
was referred to as "the year with no summer" and, more recently, as "the last great subsistence
crisis in the western world". The effects on the living conditions in Western Canada were well-
documented, and clearly placed a great deal of stress on native people and the European traders
and colonists. Friction between the groups was exacerbated by the uncertainties of food supply.
Probably the hardships created by the extreme weather were a significant catalyst for the events
that occurred.

Introduction

The effect of climate on human behaviour has been a contentious issue in the twentieth century.
The concepts that evolved from Friedrich Ratzel's Anthropo geography, published at the end of
the nineteenth century, were transported and transposed by various people until they came to a
distorted rest in Adolf Hitler's Mein Kampf. Since then the concept of climate influencing people
or history has been anathema in the academic world. Unfortunately, it is evident from even a
cursory glance at the patterns of climate and the sequence of history that we 'threw the baby out
with the bathwater'.

Department of Geography, University of Winnipeg, 515 Portage Avenue, Winnipeg, Manitoba R3B 2E9, Canada.

The following article is a precis of a public lecture given during the conference The Year Without a Summer?
Climate in 1816 at the National Museum of Natural Sciences.

196

The purpose of this presentation is not to pursue the idea of climatic determinism, but rather to
examine the pattern of the fur trade in the context of climatic conditions. The argument is
presented that plants and subsequently all animals are limited in their options and reactions by
climatic conditions. History must be examined in the context of climate because of its control
over the fundamentals of life. I prefer to think that geography and history are inseparable; history
is the play and geography the stage on which it is enacted.

It is interesting that anthropologists have little problem with the idea that primitive societies are
essentially controlled by climate, but somehow historians reject the idea. What is the difference
between the two? It is partly the fact that, until relatively recently, we have known little about
the climate of this historic period. There is also a great deal of conceit in the belief that humans
are not as affected by climate as other animals. This conceit has reached its highest levels in
North America in the twentieth century where technology is believed to have the answers to all
problems. Despite the fact that climate dictates over 80% of the yield on any farm there are no
compulsory courses in climate or meteorology at Canadian schools of agriculture. The drought
of 1988 brought the realities of the dominance of climate to the fore once again. It should have
reminded us that man's mastery over the environment is a figment of his conceited imagination.
I hope that, as we increase the amount of knowledge about past climates, we include it as a
significant factor in the pattern of human actions; both past, present and future.

The pioneering work of people like Hubert Lamb delving into historical diaries and journals to
reveal very different past climatic conditions has only occurred since the Second World War.
Historical climatology has shown that climate has varied a great deal in time and space, thus
altering the prosperity of different regions. The primary alteration is in the ability to produce
food. However, climate also affects commerce, especially if it depends upon a natural product
that is weather dependent.

One cannot examine the impact of the period from 1789 to 1820 upon the fur trade without
considering the broader context. Rarely do singular climatic periods or events create direct
change. Invariably a system is put under increasing pressure until certain climatic conditions
become a catalyst to change.

The fur trade in North America is a good example of an enterprise almost totally dependant upon
climate for its survival and success. Climate dictates: the number and quality of furs; conditions
for the trappers and their families; the ease of transport through snow conditions or water levels
in rivers and lakes; the ease of shipment across the oceans; the dependency of Europeans upon
food supply from the land, to name a few items.

There is no point in blaming historians for ignoring climate as a factor in such change because
the information has not been available. As reconstruction of climatic patterns continues, it is
essential this be included as a major factor in the mosaic of variables that direct the human
condition.

The period from 950 to 1200 is variously referred to as the medieval warm epoch or the Little
Climatic Optimum. Regardless of the term, it was a period of much warmer conditions than at
present. Oats and barley were grown in Iceland; the Domesday Book records commercial
vineyards flourishing in England; while the eleventh and twelfth centuries later became known
as the golden age in Scotland.

197

It is important to note what was happening in North America because of the parallels with current
predictions of global warming. The warmer conditions resulted in northward migration of the
agricultural people of the lower Mississippi Valley into Wisconsin and Minnesota. However, it
also resulted in increasing aridity in the Midwest - that is the area west of the Mississippi.
Analyses of Holocene pollen from the northern plains of Iowa indicate increasing aridity and a
change from deciduous forest to grasslands. In Canada the northern limit of trees expanded
northward up to 100 km in some regions, as warmer conditions brought a longer growing season.

After 1200, global climate began to cool. The circumpolar vortex expanded and the zone of
cyclonic storm activity shifted south. Cultures that had benefited from the warmer conditions now
saw a decline, but as with any climatic shift, others gained. For example, the Old Norse colony
in Greenland collapsed as crop failures increased and permafrost returned. The settlements in
Iceland and Norway experienced a decline in population as agricultural conditions deteriorated.
In North America the increased strength of the westerlies resulted in a greater rainshadow effect
in the lee of the Rocky Mountains and increased dryness on the Great Plains.

The problems in Europe were a litany of woes for people who had experienced the warm
conditions of the Little Climatic Optimum. The woes included: increasing storm severity; harvest
failures; abandonment of croplands and villages in higher elevations; and an increase in disease
and mortality rates. In Scotland it has been estimated that the elevation at which agriculture could
be practised lowered by 200 m between 1450 and 1600. The greatest loss was in the Highlands
because the vertical loss converts into a substantial horizontal loss, which is devastating in a
country with little level or arable land.

Martin Parry has estimated that harvest failures occurred one year in 20 in the thirteenth century.
By the late seventeenth century this had been reduced to one year in two. Consecutive years of
failure led to consumption of seed grain, thus accentuating the situation. The implications are that
the initial Highland clearances were caused by climate, not by land-hungry landlords. With
Highland clans forced to lower ground, the clan wars began. This was to be the beginning of
many decades of extreme social upheaval. In 1675 the Philosophical Transactions of the Royal
Society reported that a lake in Strathglass had "...ice on it in the middle, even in the hottest
summer." It is also reported by others that there was permanent snow on the tops of the
Cairngorms. We rarely stop to think that curling, which originated on lake ice, could not be
played in many winters in the twentieth century.

It has been estimated that by 1691 over 100,000 Scots had been transplanted to Ulster, driven by
such conditions as occurred in March 1674 when excessively heavy snows, severe frosts and
13 days of drifting snow resulted in the deaths of hundreds of sheep. Unfortunately the 1690s
brought even worse conditions; in the eight years from 1693 to 1700 there were seven failures
of the essential oats harvest. An excellent measure of the degree of cold during this general
period was the winter of 1683, known as the year of the great frost, when 2 feet (0.6 m) of ice
formed on the Thames River in London.

Hubert Lamb called 1450 to 1850 the Little Ice Age. The coldest portion of this period was from
1645 to 1715, with the nadir occurring in the 1690s. (Ironically this period coincides closely with
the lifespan of the astronomer, Edmund Halley). The impact of these climatic conditions on
Europe are receiving growing attention. A very important point is that this 70-year period
coincides with a period known as the Maunder Minimum. During this time there were virtually
no sunspots. There is increasing evidence that when there are high numbers of sunspots, as in
the 1980s, the Earth is warm, and when there are few it is cold.

198

Although the initial hardships of the Little Ice Age were caused by deteriorating weather, the
reaction of the landowners was, in most cases, reprehensible, ranging from absolute cruelty to
benign neglect. One who did attempt to alleviate the problems faced by some of his tenants was
Thomas Douglas, 5th Earl of Selkirk. It is essential to understand that even his motives are
suspect. That is they were not as altruistic as people have proposed. His objective was to ensure
colonization of the North American continent to halt expansion of the American revolutionary
nation. He detested revolutionaries, but especially Americans. The family house had been
attacked by John Paul Jones, Scottish-born naval hero of the American War of Independence.
Young Selkirk - then seven years old - was so frightened and angered that he held a lifelong
grudge against Americans.

Climate and Fur Trade in Western Canada

I will return to Lord Selkirk later. It is necessary now to look at the evolution of the fur trade.
I believe it is significant that the Hudson's Bay Company1 received its charter in 1670, just 10
years before one of the coldest decades in the last several hundred years. The demand for furs
would have been much different in the warmth of the Little Climatic Optimum. The Company
prospered as the demand for furs increased, and they expanded their operation accordingly.
Interestingly they did not move inland from the shores of the Bay remaining, as Edward
Umfreville described it, "asleep by the frozen sea." This was to place increasing pressure on the
wildlife in the northern regions, which will be discussed later.

The first half of the eighteenth century saw a gradual growth of the fur trade. A widening region
yielded more and more furs, but now there was a growing confrontation. The pedlars or
Canadians (as the Hudson's Bay Company called the fur traders of the North West Company
operating from Quebec) expanded westward across the prairies. Debate began within the
Company about the need to open inland posts to offset the threat. Historians argue that the
decision to move inland was totally due to this competitive factor. I contend that competition was
a factor, but more significant was the impact of climate, especially in the northern regions around
and west of the Bay.

While the debate was occurring, global climate was changing. The weather records for Churchill
and York Factory show a significant shift in the pattern of winds, precipitation and other
variables. Prior to 1760 the mean summer position of the Polar Front was south of Churchill and
York Factory. This meant that both sites experienced subarctic climatic conditions. After 1760
the Front shifted north, so that Churchill continued with a subarctic climate but York Factory
now had a more temperate climate capable of supporting the boreal forest. By the 1780s another
shift in climate was occurring. Conditions deteriorated and the record becomes replete with
comments on the lack of game, hard times and starvation among the Indians. At the same time
there was a continued reduction in the number of furs being taken.

A study of the period from 1780 to 1800 by Stephen Wilkerson quantified the deteriorating
conditions. Content analysis (a frequency count of the number of references to starvation and
other key words) clearly shows a system under extreme stress. In some years, such as 1792 the
number of comments about lack of food, starving people, malnutrition, and death are eight to 10
times above previous periods. An entry for 17 January 1792 reads, "Indeed this winter has been

The body of water is named Hudson Bay: the company is correctly called "The Hudson's Bay Company" or
sometimes just "the Company".

199

so far the most remarkable for scarcity of provisions for neither Englishman or Indians can find
anything to kill."

These conditions are coincident with unusual patterns of animal behaviour. Joseph Colen recorded
the following in the Journal for 22 October 1787, "Game of all kinds scarce but that White Bears
are so numerous and trouble some as to attack them and their stages where their provisions is
deposited." Later in the same year an entry for 23 December reads, "Late in the evening large
herd of wolves surrounded the Factory." Both these events are unusual for the period. Colen was
to become a victim of the conditions.

During this period the number of furs taken was significantly reduced. Colen wrote to London
arguing that the cause was overlapping. The Company accused him of mismanagement and
removed him from his post. Actually he was right, but for the wrong reason. The number of furs
taken would not normally have been a problem except that now the climatic changes had altered
the thresholds. Wide variations in temperature created great stress on the plants and animals. For
example, an entry for 5 February illustrates the unusual nature of the situation, "They (Indians)
also inform me that the winter set in so early upwards that many Swans and other waterfowl were
froze in the Lakes and they found many of the former not fledged, they likewise say that the
snow is remarkably deep."

However variations in precipitation caused the greatest difficulties. An entry for 14 November
1783 reads, "Never remember the snow so deep at this season of the year." The next year an
entry for 24 April informs that "Snow at least 10 feet deep." Too much snow creates great
difficulty, especially for larger animals including man. Too little snow is devastating. Ptarmigan,
lemmings and many other species that form a major portion of the base of the food chain die off
without the insulating effects of snow. Low temperatures that would not have been a problem
with deeper snow became deadly.

Expansion to the interior continued apace, and by the turn of the century competition between
the two companies was placing even greater pressure on the resources. The Indians were caught
in the middle of the conflict. They watched the battle and felt the effects as the land and its
resources were hard hit. For 20 years the struggle continued, finally being resolved with
amalgamation in 1821.

Climate improved briefly in the first decade of the nineteenth century, but by 1809 a cooling
trend was beginning. Much has been written about 1816, the year with no summer. It was
originally thought that the eruption of Tambora in 1815 was the cause of this dramatic, history-
altering year. However, climatic records show that cooling had begun in 1809, and the volcanic
eruption occurred at the nadir of the cool period 1809-20. It is interesting to speculate on the
impact of the volcano if global temperatures had been increasing at the time.

Conditions in Europe had been very similar through the latter part of the eighteenth century, as
the work of many scholars attests. Harsh conditions seriously reduced the food supply and placed
populations under increasing stress. People responded in their traditional ways, starvation and/or
migration. Interestingly, migration is not the choice of the majority. Lord Selkirk's offer of
transport to new opportunities and better conditions made the decision a little easier; but still it
was not everyone's choice.

The first groups who came from Scotland under his auspices went to Prince Edward Island and
Ontario. The best known group was the Selkirk Settlers: under the leadership of Miles Macdonell

200

they arrived at York Factory in the autumn of 1811. Too late to travel south, they wintered at
a place known as the Nelson Encampment and experienced the harshness of conditions of that
part of the world. Things were not much better when they arrived at the Red River settlement
at the junction of the Red and Assiniboine rivers in 1812. The cooler conditions discussed above
had already begun, and made the early years very difficult. The harsh conditions of the "year
with no summer" were just the beginning. In his district report for 1819 Peter Fidler writes that
from 1816 to 1819 a severe drought affected everything and everyone:

The spring months have sometimes storms of wind and thunder even so early
as March within these last three years the Climate seems to be greatly changed
the summers so backward with very little rain and even snow in winter much
less than usual and the ground parched up that all small creeks that flowed
with plentiful streams all summer have entirely dried up, for these several
years loaded craft could ascend up as high as the Elbow or Carlton House but
these last 3 summers it was necessary to convey all the goods from the Forks
by land in Carts...

The latter comments refer to the shift from river traffic on the Assiniboine River to the use of
Red River Carts. This climate-induced shift is reflected in the pattern of roads and settlement
across the prairies today.

Consider the situation that has developed by 1817. The Selkirk Settlers have been thrust into a
new harsh landscape. They are almost immediately confronted with severe weather that made
things as difficult as they had been in their native Scotland. In addition, they were not welcome,
either by the Indians, the M6tis, or the fur traders. The Indians saw them as a threat to their
traditional way of life and usurpers of their land. They were particularly concerned because they
were suffering severely by the harsh weather and consequent lack of food. The M6tis were
already the 'in-between' group and had nothing to gain from the addition of another faction.
Besides, they also benefited as a key part of the fur trade. Fur traders saw the settlers as a threat
to their freewheeling monopolistic style. They also recognized that agriculture and fur trapping
were potentially mutually exclusive. Amalgamation between the two companies had not occurred
yet, and the settlers were unwitting pawns in the conflict.

The entire situation was extremely volatile. It is not surprising that the Indians and M6tis attacked
the settlers or that the fur traders (particularly members of the North West Company) did little
to assist them. The culmination of these conflicts was the Seven Oaks Massacre in 1817 when
a group of M6tis led by Cuthbert Grant killed 21 people. I do not think that the climate was the
cause of this event, however, I suggest that the extreme climatic conditions and their impact on
the economy and food supply created untenable and volatile situations. The fur trade was to
continue for some decades as there was little useable land in northern Canada. However, the
southern regions were irretrievably changed as the land was cleared at an increasing rate.

Conclusion

Borisenkov, the Soviet climatologist, has carried out extensive research using historical sources
to reconstruct climate over the last several centuries. He writes that, "In the climatic sense the
Little Ice Age was highly variable both spatially and temporally. The main feature of that period
was the frequent recurrence of climatic extremes, during which Russia suffered 350 "hungry
years" as a result of unfavourable climatic conditions." He makes links between the climatic
disasters, such as drought, rainy and cool summers or severe winters and the pattern of peasant

201

life. The correlation between particularly prolonged harsh conditions and peasant revolts cannot
be ignored.

It is easy to blame historians for not considering climate as a major factor influencing important
social events. They have not had the information - although some clues should have been
apparent. Paintings such as those by Breughel showing winter conditions very different than today
or Jan Griffier's frosty painting of the River Thames with 2 feet (0.6 cm) of ice in the great frost
of 1683, cannot be accused of artistic licence. As the picture of historic climate is reconstructed
we should be able to reach more precise conclusions about the relationship between climate and
history.

The concept of climate influencing the pattern of history has suffered from the extreme distortions
of fascism and the lack of information about actual climatic conditions. Climatic determinism is
not the issue here, especially as it relates to human characteristics. My point is that climate affects
the environment which has direct impact upon the food supply and economy, and therefore the
people.

The fur trade of North America provides an excellent opportunity to study the relationship
between climate and the pattern of history. Records maintained by the Hudson's Bay Company
provide detailed evidence of the climate and its impact on the economy and lives of the people.

202

The Ecology of a Famine: Northwestern Ontario in 1815-17

Roger Suffling1 and Ron Fritz1

Abstract

The extraordinary summer weather of 1816 has been blamed on the eruption of the Tambora
volcano in 1815, and has been associated with various social and economic disruptions around
the globe. In northwestern Ontario there were famines in the winters of 1815-16 and 1816-17
among native Ojibwa people and Hudson's Bay Company traders who relied on seven basic
resources: moose, caribou, wildrice, potatoes, fish, wildfowl and furbearers. The first of these
two famines was extremely severe. It resulted from an initial, non-climatically induced reduction
in moose and caribou, and a natural cyclic crash in the snowshoe hare population. The early
summer drought of 1815 reduced the potato crop, and late summer rain and cold ruined the
wildrice harvest and fishing. These, combined with an early fall goose migration left both the
Ojibwa and Hudson's Bay Company employees starving. In contrast, the dry cold summer of
1816 fostered production of a large potato crop, though of indifferent quality, and normal
wildrice and fish harvests. The winter 1816-17 famine resulted primarily from deep snow
conditions, but the Hudson's Bay Company was able to feed many starving natives. Thus the
cold, droughty summer of 1816, the "year with no summer", may have done as much to
ameliorate famine among the Ojibwa people as to create it.

Introduction

Northwestern Ontario (Figure 1) is a harsh land of subarctic forests. Even now, the population
is sparse and, in the early nineteenth century, it probably never exceeded a few thousand (Bishop
1974). Though severe epidemics of smallpox and other diseases periodically devastated the
Ojibwa and Cree peoples (e.g., Hearne 1791), starvation was often the most effective controlling
factor, as the following report from Osnaburgh House (51° 90'N 90° 15'W) illustrates:

"An Indian woman of the Crows gang came in to [sic] day with her four
young children all much starved and with a very miserable report. Says that
her husband starved to death two winters back since which she has been in
wretchedness and want with four children to support. Her friends take but
little notice of her being under the impression that she eat her husband when
under one of the greatest of all miseries extreme starvation." (1 November
1814)2

The above is part of a daily record kept by Hudson's Bay Company (HBC) Post Masters, and
known as a Journal of Occurrences. It spans the years 1786 to 1911 with few interruptions. The
journals confirm the generality of the appalling conditions described above (though this is a
severe case indeed). Of 126 years of records, 60 include allusions to people starving (Figure 2).

1 Faculty of Environmental Studies, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada.

2 Unless otherwise noted, dates quoted are from the Osnaburgh House Journals of Occurrences, originals of which
are kept at the Hudson's Bay Company Archives in Winnipeg, Manitoba, Canada.

203

28 0 28

Lac Seul

-A

LAC SEUL
HOUSE

- ESCABACHEyAN
HOUSE

108 kiv^Hudsofw

Lowlands^ I

mm me

OSNABURGH
HOUSE

Hudson |Ba y ^ -I ir

fii h any 'R< ri

86c

82!

78;

Figure 1: Locations mentioned in the text.

In using the word, "starving", we intend the same meaning attached by the Post Masters: that
somebody was unable to procure food for days at a time. Sometimes this situation would be brief
or intermittent. On occasion it continued until death ensued. Often, however, it is apparent that
starvation contributed to death by other means. The woman cited above struggled on in the same
pathetic condition for two more years before succumbing to an illness: she had killed her husband
to survive a famine.

We have used the word "famine" to mean general starvation among people that was sufficiently
prolonged to become life-threatening.

204

250

iDomomou^omomou^omoLnoLno
i^r^r^cococorowcococococooocococooococo

YEAR

Figure 2: Incidence of starvation at Osnaburgh House. The starvation index is the product of the number
of references to starvation in the Journal of Occurrence and the maximum number of people
recorded as starving.

In the fall and winter of 1815-16 there was a particularly horrendous incident of famine in
northwestern Ontario that is well illustrated by the Osnaburgh House records. It involved not only
the native population, but also (and unusually) the better prepared HBC employees and their
families. The 1815-16 famine was one of eight recorded that involved more than 30 people (out
of a total population of about 200). In terms of numbers referenced in the journals, it was one
of the worst eight incidents, and in deaths it probably ranks only second to the 1823-25 incident.

There was another famine in the 1816-17 winter but, though it was severe, and though it may
have involved just as many people, its consequences were not as grave as those of the 1815-16
famine. Both of these famines are part of a prolonged series of incidents from 1810 to 1825 that
broke the spirit and culture of the Ojibwa people (Bishop 1974). The famine series is associated
with depletion of big game and an exceptionally cold climatic fluctuation.

The 1815-16 famine is particularly striking because it corresponds with climatic fluctuations
evidently caused by the eruption of the Tambora volcano in what is now Indonesia, early in 1815.
The question that we asked ourselves, therefore, was whether the 1815-16 famine was caused by

205

unusual weather conditions. To find an answer, we examined the ecology of the Ojibwa people
around Osnaburgh House to see which climatic or other conditions normally contributed to
famine, and to see which of these pertained immediately before or during the 1815-16 incident,
and the lesser famine of 1816-17.

Ojibwa Ecology and Food Sources in Early Nineteenth Century Northwestern Ontario
Big Game

Bishop (1974) has postulated that the Ojibwa people who lived near Osnaburgh House in the early
1800s had moved there from around Sault Ste. Marie.

Initially, they had been primarily big game hunters, subsisting on moose (Alces alces) and
woodland caribou (Rangifer tarandus). Bishop believes that, after first European contact, the
people began to make forays into northwestern Ontario in search of furbearers to use in the new
commercial fur trade. Rival companies soon began to challenge the HBC's hegemony over this
and other areas. First the French and then Scots from Montreal, and American traders moved into
the area to trade furs at their source. This induced the Ojibwa to remain on the summering
grounds year round, but forced a radical reorganization of their hunting strategy. Originally, they
had hunted in large groups. Now, with the need to spread out to trap beaver, they broke into
family groups, and they used firearms to kill moose and caribou. When they were available in
sufficient quantity, moose and caribou meat and skins were traded to the HBC, putting further
pressure on the herds. By 1815, both species were already somewhat depleted (Figures 3, 4), and
this was beginning to wreak hardship, not only directly in terms of food availability, but also
because leather for mocassins and snowshoes was becoming scarce. The people's very existence
in this land of thinly-spread resources, was predicated on nomadic foraging, so a lack of leather
hampered many food gathering activities, as well as in fur trapping. In a typical instance, a native
arrived at Lac Seul asking to purchase a summer bear (Ursus americanus) skin, there being no
caribou or moose leather: "The bear skin is for making his shoes without which he cannot leave
his tent" (Lac Seul 28 April 1828).

Moose (Figure 3) and caribou (Figure 4) were stalked at all times of the year, and herded into
the water for slaughter in the summer. Deep snow slowed the animals down in winter, making
them easier to approach, but they could easily outrun hunters on thin snow - so the latter
condition is associated with hardship. Extremely deep snow made both hunter and hunted less
mobile and sometimes prevented the people from traveling between various moose and caribou
wintering grounds.

Crusted snow gives human or other predators a marked advantage (J. Theberge 1988, personal
communication). It must have occurred with greater frequency in the early nineteenth century as
uncontrolled forest fires increased the proportion of open country where crusting occurs easily.
Thus, even as the herds were reduced, the pursuit of the remaining animals may have become
more efficient, ensuring further big-game depletion.

Moose and caribou meat were eaten fresh, or preserved by drying in strips over a fire, or in
pemmican (a preserved mixture of fat, berries and shredded meat pounded together).

206

1 5

o
o

I 1 1 1 1 1 1 1 I 1 1 I !

o in o
o o

ll 1(1 III I II I I II I 1 1 I I 1 1 I 1 1 II M I 1 1 I I II I II I I I! I I 1 1 II I I I II I 1 1 I I 1 1 I I I I I 1 1 I II I I 1 1

m o m o m o

(M N K) IO *

o m o m
in in id

in o m o in

03 03 O) Oi

r^r^i^iococoxcocococococococococococococococo

o in o
o o «-

o> W Ol

YEAR

Figure 3: The number of moose involved in trading of meat and skins from natives to the HBC at
Osnaburgh House. The upper line is a maximum estimate, and the lower line a minimum.
Derivation of the data is given in Fritz (1988).

cDomomomomomomomomomomomomo
oooioioo^-T-csicNfon^-^ininiDiDr^r^oooooicDOO'-
r^i^i^co»cocooooococococococococococococococoo5CT)cn

YEAR

Figure 4: The number of caribou involved in trading of meat and skins from natives to the HBC at
Osnaburgh House. The upper line is a maximum estimate, and the lower line a minimum.
Derivation of the data is given in Fritz (1988). The maximum figure (102) for 1876 is off-scale.

207

Fish

The third major food source was fish. They were hooked, speared or netted, depending on
species and season. Several species were used including: whitefish (Coregonus clupeiformis),
sucker {Catostomus spp.), pickerel (Stizostedion vitreum), sturgeon (Acipenser fulvescens) and
pike (Esox lucius). At Osnaburgh House, sturgeon appear to have been particularly critical to
human welfare. They could be readily speared and netted when spawning in the early spring -
an otherwise lean time of year. Spring and summer were employed in catching mostly pickerel,
pike and whitefish; and the fishery continued until the water became warm and the eating quality
of the fish declined. Fishing resumed in the fall as water temperatures fell, when a number of
species came to spawn in the shallows and rapids of the rivers. Fishing continued until freeze-up
and occasionally afterwards, under the ice, but the early nineteenth century natives do not seem
to have mastered the art of ice fishing with nets as the HBC people had.

High water in the lakes generally signalled a failure of the fishery, especially in the fall. The high
water could be caused by unusually heavy rainfall, cool weather, or a combination of both. Early
freeze-up also hurt the fishery as it cut short the spawning seasons of the fish, and they withdrew
to deeper water.

Fish were vitally important at northwestern Ontario HBC posts during winter - especially
whitefish and, as soon as the weather became cold enough to store fish, they were netted
intensively. In times of native starvation these fish were distributed to Ojibwa begging at the
posts, as long as the HBC's own supply of stored or fresh fish remained assured. The motivation
was partly charitable but hinged too on the economic need to preserve the lives and health of the
beaver trappers who were the lifeblood of the Company's activities in these parts.

Wildfowl

The fourth major native food resource was wildfowl - primarily geese. Both Canada Geese
(Branta canadensis) and Snow Geese (Chen caerulescens) were shot, as well as a variety of
ducks. At Osnaburgh House, wildfowl first appeared in April, moving north to the Hudson Bay
Lowlands and beyond on the turbulent edge of the retreating Arctic air mass (Ball 1983). If snow
lies on the coastal marshes of Hudson Bay at the time when goose eggs should be laid, a lower
proportion of females than normal actually lays eggs. In addition, average clutch-size is reduced.
Thus the cohort of young geese produced is small, as happened in 1967. Fall migrants then prove
relatively sparse, as do those birds returning the following spring. If fall came early, sending the
birds south too soon for native needs, then people had a longer time to wait between fall and
spring migrations. The people were often starving in late winter, so that the return of the geese
was awaited with eagerness by both natives and HBC men.

Wildrice

Wildrice was the only staple vegetable of the largely carnivorous Ojibwa. (The same cannot be
said of the HBC men who also grew potatoes and some lesser crops). Wildrice is an annual
aquatic grass found in slow-flowing rivers and shallow lakes (Dore 1969, Suffling and Schreiner
1979). It sets seed in late summer and is harvested in late August or early September in
northwestern Ontario. The seeds, which were fermented, hulled, dried and stored for winter use,
were a good hedge against starvation. High water in mid- to late-summer - especially rising high
water - is disastrous to the crop. Also, windstorms can scatter the grain before it is harvested.

208

Hares

Snowshoe hares (Lepus americanus), usually called rabbits in the journals, were also an important
food item. They never appear to have been a preferred food (Bishop 1974), but were snared in
hard times when other victuals were lacking. Hare pelts that were the by-product of this activity
were traded, but only commanded a minimal price at the HBC posts. Alternate freezing and
thawing in winter made rabbit snaring impossible (Lac Seul, 3 February 1825).

The snowshoe hare exhibits a remarkable eight- to nine-year cycle of population density
(MacLulich 1937, Elton and Nicholson 1942). It is notable that most of the peaks in human
starvation at Osnaburgh House appear in the year after the crash of the hare population
(Figure 5), a relationship which is statistically significant (X2, P<0.01). Thus snowshoe hare
scarcity could precipitate famine.

Furbearers

Furbearers were the means by which the Ojibwa obtained non-local commodities such as iron
knives, hatchets, guns, blankets and rum. The species trapped or hunted include marten (Martes
americana), otter (Lutra canadensis), fisher (Martes pennanti) and lynx (Lynx canadensis), but
the most important was beaver (Castor canadensis). Where they were available in large numbers,
as at Lac Seul, muskrats (Ondatra zibethicus) were also very important in total, and as beavers
were depleted, muskrats assumed an increased economic significance. Although beavers and
muskrats had the added advantage that the carcasses were edible, generally furbearers did not
contribute greatly to human nutrition. Their significance in this context is in how they influenced
the pattern of trapping and hunting of other animals.

Potatoes

As a rule, natives in early nineteenth century northwestern Ontario did not grow potatoes, though
there were a few individual attempts. This vegetable was, however, a staple of the HBC posts.
Potatoes were planted in early May at Osnaburgh, and harvested in mid- to late-October.

The potato crop was highly variable. It appears to have suffered after hot, dry summers, and was
of low quality if an early fall frost damaged the tubers. Though spring frosts were damaging to
the top growth and may have reduced the yield, they do not seem to have been as serious a
problem. Every few years, there were also damaging epidemics of "grubs" (so far unidentified).
As with fish, potatoes were given to starving Ojibwa coming to the post for assistance - at least
for as long as there was no threat of starvation to HBC employees themselves.

The Annual Cycle of Ojibwa Subsistence

The annual cycle of Ojibwa subsistence in the early nineteenth century (Figure 5) is not totally
dissimilar to the modern pattern described by Sieciechowicz (1977). The Ojibwa year can be
thought of as beginning in late September to October when natives arrived at the HBC post to
obtain their outfit for the coming winter. At this time, goods were normally obtained on credit,
a process described as "taking debt" or "outfitting". The fall fishing and goose harvest more or
less coincided with taking debt. Then, as the weather hardened, and the furbearers came into
prime pelage, there was a concentrated effort to trap - and especially after the first snows. The
amorous bull moose could be readily killed at this time as they could be called in by a hunter
using a birch bark trumpet to initiate a rival or a cow moose.

209

When the large lakes froze, usually in early November around Osnaburgh, winter began in
earnest. By now only big game, hares and occasional grouse (Canachites canadensis, Bonasa
umbellus, and possibly Pedioecetes phasianellus) were available. Since stored fish and wildrice
gave out, often about the beginning of January, people had to since rely entirely on meat and fat.
If starvation arose, it became apparent in the journal entries at this time, and it would become
more severe after the onset of really bitter weather. The starving time of winter could be
alleviated or avoided if large game abounded, if hares were present in large numbers, or if spring
came early. The converse was also true.

JUNE

Figure 5: The annual cycle of early nineteenth century Ojibwa people living around Osnaburgh House.

210

If people were not starving severely, there was a second burst of fur-trapping activity in the
relatively mild weather of late winter and early spring.

Normally in mid-April the first geese arrived, and sometimes set off what can only be described
as a hunting frenzy at the HBC posts! At Lac Seul in 1828, for instance, the Post Master gave
all his people three days off to hunt geese,"... in the hopes of their setting to work afterwards" .
Geese not only provided relief from starvation, but for the HBC men in particular, brought a
welcome rest from the six-month monotony of potatoes and whitefish.

In late winter Ojibwa appeared at the posts to redeem their debts, to socialize, and to drink. It
was a time when the HBC Post Master anxiously awaited the fur harvest, and when the extent
of any lethal starvation became apparent through the non-arrival of families from the forest.

With the break-up of ice on the lakes and freshets of meltwater in the rivers came the spearing
of sturgeon and pickerel as they spawned at the base of rapids. Bears (Ursus americanus) too
came for the fish, and could be readily trapped then, if they had not been found in their
hibernation dens and speared. They were generally too lean in the spring to provide much meat
or fat. Fishing continued throughout the spring, supplemented by game hunting, as well as by
collecting a variety of fruits and berries, and possibly birds' eggs.

In August the water became too warm for profitable fishing, but at the end of the month the
wildrice harvest began. This was another time that brought people together, and it was closely
followed by fall fishing and acquiring new outfits at the Posts.

The pattern described above is typical, but each year presented a slightly different situation, and
the resources available around each trading post differed slightly. Osnaburgh had more sturgeon,
Lac Seul had more muskrats and wildrice, etc. The trading policies of the HBC and its rivals,
the weather, as well as availability of food and furs all varied enormously over time, and have
been discussed at length by Bishop (1974).

Factors that Precipitated Starvation

The factors causing or excacerbating starvation are summarized in Table 1. Severe starvation
might be avoided if only one or two factors were unfavourable in a given year, but if several
coincided, then people would suffer accordingly. Most of the factors have been discussed above,
but the incapacitation of hunters needs comment.

Injury, sickness, or death of menfolk was a constant peril to family groups - it could deny them
access to big game. If the women and children's occupation of snaring hare was unavailable
because of a hare population crash, then starvation was bound to follow. Repeated freeze-thaw
cycles that prevented hare snaring sometimes had the same effect.

The 1815-16 Famine

January and February 1815 were fairly typical for the time of year - dry and cold. This weather
persisted into April which, coupled with north winds, kept most of the geese and ducks from
arriving. A few came, however, on 15 April - about the usual time.

211

May, too, proved very cold at first so that the Post Master remarked that the weather had "more
the appearance of March than of May" (8 May 1815). The main body of geese arrived only on
14 May, a month late. The keeper of the journal considered that the latter half of May was warm
for the time of year, and the lake ice broke only a little late.

Early June was judged to be warm for the season, but the latter half was rainy and cold. This
weather persisted into early July until it became very hot during 9-13 June, and then again from
24 July to 10 August. The warm spell was sufficiently drying that the writer states:

"Hookamarshish informs me that all his furs were burned, he says that he was
going to move to another place and he forgot to put out his fire and so it set
fire to the woods and burned all his furs." (15 August 1815).

Table 1: Factors Contributing to Starvation among the Early Nineteenth Century Ojibwa around
Osnaburgh House.

Causal Factor

Resource Affected

Climatic Contributors

High water in summer
Rising water in summer
Thin snow
Very deep snow

Cold, damp spring on Hudson Bay

Droughty summer

Freeze/thaw in winter

Non-Climatic Contributors

Increased forest fires due to
fur trade

Overhunting due to fur trade

Cyclic population crash

Injury, sickness or death
of menfolk

Wildrice, fish
Wildrice
Moose, caribou
Moose, caribou
Geese
Potatoes
Snowshoe hare

Moose, caribou

Moose, caribou
Snowshoe hare
Moose, caribou

212

On 22 August came a sudden cooling with rainy, stormy weather accompanied by NW and E
winds. These conditions persisted unabated until 17 September. By late September it was apparent
that both the fall fishing and the wildrice harvest had failed on account of high water in the lakes
and rivers:

"The water being so remarkably high at this place the Indians is not made
any rice worth while so that I have only got 64 gal in all. So that I am much
afraid of starving in the winter as there is no fish to be got here when the
water is high in the fall. Am sorry to inform you that this is a very poor
place for most everything. There is no beaver nor moose to the indians to
hunt and most of them were starving when I seed them but are all off now
to hunt." (Letter from James Slatter at Escabachewan 23 September 1815 to
the Master of Osnaburgh House).

At Osnaburgh, the problems were compounded by a lack of fishing twine and of available labour.
Such fall starvation was unusual, but the people were probably cheered by the early arrival of
the bulk of the fall geese on 1 October. In reality, this worsened matters for, with the early
passage of wildfowl, the impending winter starvation was to last longer.

The potato harvest at Onsaburgh House was 77 kegs, down 20 from the previous year, so that
the HBC people entered the winter with very little food to spare for visiting Ojibwa.

The first snow came on 22 October and the lake froze on 7 November, a trifle early. The
subsequent ice fishery failed as miserably as had the fall netting. The first half of December was
very cold and the the latter half mild.

Starvation is first mentioned again in the journals in December, and by late January 1816 it was
general among the natives, even appearing at the HBC fishing outposts: "The men are already
feeling the iron hand of want." (Osnaburgh House, 30 January 1816).

In February, which was cold even for that time of year, natives arrived at the Post both frozen
and starved; but others coming from the north were heavily laden with furs and apparently well
fed (according to Bishop (1974) there were still moose to be had in that quarter).

By 23 March, the potato ration for HBC people had been cut to two gallons per week (instead
of the usual three three gallons), and by 25 March everybody was sent out to hunt or fish since
the daily ration was, by that time, one small fish. The men at the marsh outpost of Osnaburgh
were now too weak even to go to the House for food and one fellow, reduced to eating fish offal,
became very sick.

Very few natives had visited the Osnaburgh House during the winter, either because they were
too weak to travel or possibly because word was out that there was no food to be had there.

April was very cold, and the digging of the potato garden at Osnaburgh House began two weeks
late (on 30 April) as a direct consequence. The famine finally broke with the arrival (three weeks
late) of the first geese on 4 May. Sturgeon did not begin to spawn until 29 May - two weeks later
than usual.

213

June 1816, likewise, was very cold with a hard frost on 4-6 June and another on the 23-27 June.
On both occasions the gardens were badly frosted. On the latter, there was one-quarter inch of
ice in the bottom of the canoes pulled up on the shore. This must have been a dry month as the
lake fell six inches in three weeks.

July was cool and rainy with mostly NW winds, and this weather continued into the first half of
August. In spite of this, the water remained low in the Albany River, suggesting perhaps that
there had been little snow in the previous winter, and that water in the marshes must have been
low all summer.

On 18 August, there was snow - an unheard-of event in this month, and with continuing cold
weather the geese were already flying thick by 15 September - a month early. During 25-30
September there was a gale, remarkable not so much for its ferocity but for its duration. Its winds
tracked from E to S, to SW. In contrast with 1815, there is no indication that the wildrice harvest
or the fall fishing were other than normal at Osnaburgh House.

The ground froze by the 3 October (about three weeks early), so the HBC people were caught
unprepared and the potatoes were frozen in the ground. In spite of this, they harvested 190 kegs,
a large crop, though evidently of indifferent quality on account of the frost. The rye plants were
six-feet tall but the grain was still green, and never had a chance to ripen. The whole crop was
lost.

The lake froze a little early on 9 November and the weather continued cold until the second half
of December which proved mild and snowy.

January 1817, and the rest of the winter, were cold with heavy snowfall, which was in marked
contrast with the previous cold, and apparently dry winter of 1815-16.

There are almost as many citations of starvation in the 1816-17 journal as there had been for
1815-16, but there is little indication of the grinding life-threatening severity of famine which had
overtaken people in the previous year. Apparently the 1816-17 starvation touched only the native
people, and many of them were visiting the post for handouts.

Discussion and Conclusions

It would be easy to rush to the conclusion that the Tambora eruption of 1815 explains any
unusual weather patterns in the subsequent couple of years. As the discussions of the "Year
without a summer? Climate in 1816" conference demonstrated, it is difficult to unequivocally
establish causal connections, even though there is a suspicious conjunction of climatic dislocations
around the globe. At Osnaburgh House, the unusual conditions were: the sudden cool, wet end
to the hot, droughty summer; the long, dry, cold winter of 1815-16; and the dry cold summer
of 1816.

Likewise, the mere existence of famine in 1815-16 is not proof, per se, of the human ecological
consequences of the Tambora eruption, or even of the effects of the harsh weather of these years.
In reality, several factors contributed to the famine at Osnaburgh House and elsewhere in
northern Ontario. They include the following ecological factors:

214

1. Depletion of moose and caribou herds by overhunting (Bishop 1974), or possibly by habitat
change through forest fires in the late 1700s. Both of these are associated with the expansion
of the European fur trade.

2. A cyclic crash in the hare population between spring 1814 and spring 1815 fur returns.

Neither of these two factors are climatic, but they are primary causes in the sense that they set
the stage for the other events. The fate of many natives was sealed by other phenomena that were
indeed climatic, namely:

1. Failure of the wildrice harvest in 1815 due to high, rising water levels in late summer.

2. Failure of the 1815 fall fishery due to high water, and failure of the ice fishery, possibly for
the same reason.

3. The small 1815 potato harvest resulting from the early summer drought and hot weather. (The
potatoes were grown in a dry, "hungry", sandy soil that warmed quickly in spring but was
very vulnerable to drying.) Thus the HBC had few or no potatoes to spare for the natives
during the famine.

4. The early-fall goose migration in 1815, and the late-spring migration of 1816. The former
ensured that the natives entered the winter in poor nutritional condition, and the latter
prolonged their suffering in spring.

One last factor was the lack of HBC labour for fishing, and lack of fishing twine in the fall 1815.
These are minor economic or logistic causes.

The summer of 1816 was in marked contrast to that of 1815. Though there were late spring frosts
and summer snow, the cool weather actually appears to have helped the potato crop. Likewise
the droughty conditions kept water levels low and assured at least a normal fish and wildrice
harvest. Spring frosts had few harmful effects, and the fall frosts did some damage to potatoes,
but not enough to be serious.

It is as impossible to say that climate alone caused the 1815-16 famine as to claim successfully
that non-climatic factors were responsible. People relied on seven major resources. Moose and
caribou had already been depleted by 1810, and the worsening climate of 1810 to 1817 was the
trigger for a series of famines that were only alleviated by the hare "high" of 1814. The
subsequent crash of hares reduced the resources available to four: wildfowl, wildrice, fish and
potatoes. The high water of 1815 knocked out two of these - fish and wildrice - leaving only
potatoes and wildfowl - and even the potatoes were reduced by early drought. A catastrophe then
became inevitable.

The fall of 1816 was different. Big game and hares were still scarce, but the water remained low,
ensuring a wildrice and fish harvest. The cool summer evidently favoured the HBC with a large
potato crop, although possibly more had been planted as a reaction to the previous famine. On
the other hand, the October frost impaired the quality of the crop. The two frosts of June had
evidently had little effect, although one might easily jump to the conclusion that they had caused
the smaller famine of 1816-17. In reality the potatoes were the saving grace for a native
population that probably was still reeling from the physical and psychological impact of the
1815-16 famine. The winter or 1816-17 was a year of starvation, but was not as serious as its

215

predecessor. If anything, it must have been deep snow that limited travel to new hunting grounds
or to the HBC for charity, that caused most hardship.

We conclude that the extraordinary cold, dry weather in the summer of 1816 may actually have
done more to prevent famine than to create it. The cold, wet end to the summer of 1815 was,
however, the proximal cause, but only the proximal cause, of the 1815-16 suffering.

The postscript of the famine is as interesting as the event itself. Between 1819 and 1820, the
number of both moose and caribou traded at Osnaburgh House rose dramatically (Figures 3, 4).
Perhaps enough hunters perished that the predation pressure on the herds was reduced and they
started to increase. Deep snow in 1816-17 may have reduced hunting pressure with the same
effect. Whatever the causes, moose and caribou then persisted until the mid- 1820s before
succumbing to hunting or other pressures. Thus one effect of the 1815-16 incident had been to
prolong the survival of big game that were so important to the people, even as it helped to
destroy their will and culture.

Acknowledgements

The research for our paper was conducted while one of us (R.F.) held a Natural Sciences and
Engineering Research Council Undergraduate Internship. We thank John Theberge and
Harold Lumsden for their advice concerning ungulate and goose ecology.

References

Ball, T. 1983. The migration of geese as an indicator of climate change in the southern Hudson
Bay region between 1715 and 1851. Climatic Change 5:85-93.

Bishop, C. A. 1974. The Northern Ojibwa and the Fur Trade, An Historical and Ecological Study.
Cultures and Communities Series. S.M. Weaver (general ed.). Holt, Rinehart and Winston,
Toronto.

Dore, W.G. 1969. Wildrice. Canada Department of Agriculture Research Branch, Plant Research
Institute Publication 1393. Ottawa. 84 pp.

Elton, C.S. and A.J. Nicholson. 1942. The ten year cycle in the numbers of the lynx in Canada.
Journal of Animal Ecology. 1:215-244.

Fritz, R. 1988. Moose and caribou population decline in N.W. Ontario boreal forests of the
Osnaburgh House (HBC) trade area: 1786-1911. Senior Honours Essay. Department of
Geography, University of Waterloo, Ontario.

Hearne, S. 1791 . A journal of observations made on the journey inland from Prince of Wales Fort
in latitude 58° 50' North to latitude 72°00' Beginning 7th Deer. 1770, ending June 30th,
1772 by Samuel Hearne. Manuscript in British Museum Library, London, U.K.

MacLulich, D.A. 1937. Fluctuations in the numbers of the varying hare (Lepus americanus).
University of Toronto Studies, Biology Series 43:1-136.

216

Sieciechowicz, K. 1977. People and land are one: an introduction to the way of life north of 50°.
Bulletin of the Canadian Association in Support of Native Peoples 18(2): 16-20.

Suffling, R. and C. Schreiner. 1979. A Bibliography of Wildrice (Zizania species) Including
Biological, Anthropological and Socio-economic Aspects. University of Waterloo School
of Urban and Regional Planning, Working Paper 5, Waterloo, Ontario.

217

The Development and Testing of a Methodology for Extracting Sea-Ice
Data from Ships' Log-Books

Marcia Faurer1

Abstract

The severity of the weather in 1816 in the Hudson Strait and Hudson Bay regions became
apparent through the reconstruction of sea-ice conditions for the period 1751-1870. The sea-ice
data were derived from an exceptionally large collection of ships' log-books. Current research
is focusing on the development of a reliable methodology for use in further environmental
reconstructions covering this period using historical documents.

This research applies a methodology called content analysis which was developed by the Social
Sciences for extracting meanings from human communications in an objective manner. This
technique has the ability to reduce the subjectivity inherent in the interpretation of historical
documents by testing the level of reliability of the procedure that is used to extract sea-ice data
from log-book descriptions. In this study, tests have been applied throughout the development of
the methodology with the goal of devising an objective procedure. These tests also reveal the
degree of detail that a particular source can reliably provide, as well as helping to reduce the
difficulties associated with calibrating the historical terminology against the contemporary sea-ice
vocabulary.

Introduction

Although content analysis (CA) has been used widely in the application of historical documents
as proxy sources for climatic reconstructions, this methodology has not been applied to its fullest
potential. This is primarily due to the general omission of its strongest attribute, which is the
ability to test the reliability of the methodology that is used. This aspect of CA is not merely an
option that may or may not be applied, it is actually an integral part of the CA process. Without
this means of evaluation, the interpretation of historical texts may be guided by predetermined
decisions about the information required for the reconstruction instead of by the information that
the documents can objectively provide.

This case study was conducted to test the applicability of an objective methodology for extraction
of environmental data from historical documents. The format of CA was followed closely by the
repeated application of reliability tests. Whereas this reduced the information obtained, it insured
that the derived data were obtained to a measured and acceptable level of reliability.

Data Sources and Background Information

The eighteenth and nineteenth century log-books of the Hudson's Bay Company are a potential
source for a wide variety of environmental data. Although temperature readings were entered in
the log-books, they appeared sporadically throughout the period of record (1751 to 1870). Wind
directions were given on a fairly regular basis as well as other meteorological phenomena,

1 Department of Geography, University of Manitoba, Winnipeg, Manitoba R3T 2N2, Canada.

218

however sea ice was chosen to be the focus of this study. While sea ice is not a meteorological
element per se, it is a visible expression of several environmental factors. It was also selected
because it posed a clear and present danger to the success of the voyage and to the lives of the
crew. Therefore, it was anticipated that any event or observation related to sea ice would be
faithfully recorded in the log-books.

Sea ice was encountered by the ships in Hudson Strait and Hudson Bay during the westward
portion of a voyage between England and the Hudson's Bay Company's bayside posts. These
locations and the routes of the ships are shown in Figure 1. Each year, the Company dispatched
a small convoy of ships to supply these remote posts and to bring trade goods back to England.
This collection of log-books provides an unbroken record of sea-ice descriptions that can be
cross-checked because each of the ships in the convoy kept at least one log-book. Figure 2 is a
reproduction of a log-book page showing the meticulous way in which the environmental
observations were recorded. Fortunately, this format and the vocabulary used in the log-books
remained virtually unchanged throughout the entire period of record.

Figure 1: Sea currents and Hudson's Bay Company sailing route.

219

/A /■

fin

A /

A-

/
a

p
o

/ry/7, ) 7"

as •

its

A-

jr.

A-

/LJs i

7 S

4 ^ S/L^jL.

* / -0

' k £j4* n'.T* ~»

~4- ■

Figure 2: Sample log-book page.

220

Problems of Interpretation

The problems arising from this type of data source are twofold. The first aspect of the problem
lies in the interpretation of the descriptive accounts that used a vocabulary which was different
from the current terminology. Secondly there is the difficulty that arises in conversion of
qualitative or verbal descriptions into quantitative or numerical data that can be compared with
contemporary records. When properly applied, CA serves to resolve these problems reliably.

Although these historical sea-ice descriptions used the same terminology throughout the period
of record, their conversion into a sea-ice index was complicated by two factors. The first is that
the Hudson's Bay Company did not provide a dictionary of the terms since they were not
originated by the Company but were passed down through two generations of ships' captains.
Therefore it is not possible to translate the terms directly from an historical lexicon to their
current counterparts. Secondly, even though the individual words may have been consistent
through time, they were not used in isolation, but rather they were used in phrases and sentences.
As a result, their meanings changed in relation to the context in which they were used. An initial
survey of the log-books resulted in a compilation of 90 representative phrases, an example of
which is given in Table 1.

Consequently, the system of classifying these phrases had to allow for any number of
combinations of terms. In cases where there is a finite set of terms or phrases, it is possible to
establish specific rules and guidelines for their classification. In this case however, this approach
would not have been sufficiently flexible. As a result, it was obvious that a certain degree of
subjectivity would be an unavoidable component of the analysis. The goal therefore, was to
devise an objective system of categories.

Methodology

Figure 3 illustrates the CA procedure in the role of a translator from the written descriptions into
numerical ice data. Even though the specifics of CA are usually designed to suit the needs of each
individual research project, there is an established general format (Figure 4).

Qualitative Description Derived Quantitative Data

Figure 3: Conversion of qualitative data to quantitative data through content analysis.

221

CONTENT ANALYSIS PLAN

COMMUNICATIONS
CONTENT
Log-Books

UNACCEPTABLE
RELIABILITY

RESEARCH QUESTION

CATEGORY
DEVELOPMENT

CODING SYSTEM

RELIABILITY
TEST

THEORY

Sea Ice Processes
in Hudson Bay

ACCEPTABLE
RELIABILITY

CODING

RESULTS
Derived Sea Ice Data

VALIDITY
TEST

Figure 4: Content analysis plan.

222

Table 1: Representative Eighteenth and Nineteenth Century Sea-Ice Phrases.

1. Ice open and heavy

2. Ice close but much smaller

3. Sailing among heavy straggling ice

4. Pieces of ice

5. Passing thro' a deal of sailing ice

6. Passing thro' close ice

7. Heavy close packed ice

8. Saw some ice

9. Fast beset among close small ice can't move

The formulation of the research question is the first of a series of decisions that are made
throughout the procedure. The crucial nature of this decision is due to the fact that CA is a linear
process in which each step is the logical outcome of the previous step. Should an error be made
anywhere along the line, then it will be carried throughout the entire analysis and may even be
intensified in the process. The research question is based on two sources of information: the
communications content and a body of theory. In the case of this study, the ships' log-books of
the Hudson's Bay Company provided the communications content, and the Hudson Bay sea-ice
processes provided the background theory.

The set of procedures that follow from the identification of the question (Figure 4) comprises the
core of this research. The steps followed here are repeated until an acceptable degree of reliability
has been attained. This means that a group of independent researchers who apply the same
methodology using the same data will consistently produce the same results. Basically, this is a
process of redefining categories with the goal of reducing the level of subjectivity. Once this has
been accomplished using a sample of the data, the categories can be applied to the entire body
of descriptions, and the reconstruction can be made and tested for validity.

This research followed the plan presented in Figure 5, and was divided into three phases. The
goal was to objectively develop a set of categories into which the log-book descriptions could be
grouped.

Phase I

The first phase involved 56 randomly selected log-book pages that included samples from the
120-year period, and five coders who all had considerable experience with CA and the log-books.
This phase was essentially experimental. A set of five categories and codes (Table 2) was
intuitively derived, and was based on the maximum amount of sea-ice information that was

223

RESEARCH PLAN

PHASE I

- 5 Categories : Intuitively derived code

- 3 Coding Units : Day, Entry, Word

- No Coder Training

I ~
Coding

Reliability
Tests

PHASE II

- 5 Categories : Code derived from

Observers' Manual

- 1 Coding Unit : Watch

- Definitions & Diagrams Provided

Coding

Reliability
Tests

Code

Modification

PHASE III

4 Categories : Code derived from

Phase II
1 Coding Unit : Watch
Definitions, Diagrams, Photos

Tests Confirm
Modification

Reliability
Tests

Figure 5: Research plan.

224

desired for the reconstruction. Each of the five major headings required the coders to make a
dichotomous decision about the meaning of the log-book transcription. Besides the 10 codes, the
coders were given the option of indicating those transcriptions that did not provide enough
information to make a decision regarding the particular category. This was an important aspect
of the code insuring that each decision made by the coder was done with some degree of certainty
and was not a forced response. It also allowed decisions to be made later about the type of
information that could be obtained from the log-book descriptions.

The first phase of coding was actually divided into three sections based on the unit of transcribed
information that was coded (coding unit). In the first section, the coders were required to provide
a five-digit code (one from each of the five categories - see Table 2) for each of the 56 days. In
this way, a day was treated as one block of information or coding unit. This process was repeated
by each coder five times so that their level of consistency could be determined. The second
section involved the coding of each individual hourly entry for the same 56-day sample (a total
of 261 entries), and again this process was repeated five times. The third coding unit was
comprised of a list of 81 individual words, 24 of which were direct descriptions of sea ice, and
57 described the navigational activities employed to deal with the ice (e.g., grappling, tacking,
rounding). These words were only coded twice.

To evaluate the reliability of the code, the coding units, and each coder, percentage agreements
were calculated. This is the most common and rudimentary method of calculating reliability. It
is a considerably less-than-ideal approach because it is biased in terms of the number of
categories and coders, in such a way that the fewer of each the higher the percentage agreement
is likely to be. The highest average intercoder agreement (among the coders) was 53.8% which
was achieved when the day coding unit was used. The highest intracoder agreement (consistency
level for each coder) was 80% for the entry coding unit, although all three showed high levels
of consistency (day '=68% word =70%). It is important to note however, that a large proportion
of the agreements was due to agreements that there was not enough information.

Phase II

The second phase was derived only slightly from the findings of the previous phase. Again there
were no compulsory categories so that a coder could judge that the information was insufficient
to make a coding decision. As a result of this option in the first phase, it was possible to conclude
that sea-ice concentration was the only category in which a decision was possible as much as 70%
of the time.

Phase II introduced a new coding unit: the seamans' watch, which was more in context with the
log-book format since the entries were actually summaries of the six four-hour watches listed in
Table 3. Therefore the coders were required to provide a code for each watch per day (whether
there was an entry or not). Another difference between the two phases was that the second code
was not intuitively derived nor did it evolve from the first coding system. Instead, it was based
on the terms and definitions found in The Ice Observer's Training Manual (Environment Canada
1984). Since the final goal was to create a sea-ice reconstruction that could be compared with
current records, the logical approach was to use modern definitions in developing the categories
and codes, and in the coding process itself. The second set of categories and codes is given in
Table 4, and with this the coders were also given definitions and diagrams (from the Observer's

225

Table 2: Phase I Code.1

Presence

0 = Ice not present in vicinity of ship

1 = Ice present in vicinity of ship

Concentration

2 = Small area covered by ice (<50%)

3 = Large area covered by ice (>50%)

Fragmentation

4 = Ice cover highly fragmented

5 = Ice cover not highly fragmented

Thickness

6 = Thin layer of ice

7 = Thick layer of ice

Motion

8 = Ice in motion

9 = Ice not in motion

1 (No compulsory codes). Coding units: Day (5x), Entry (5x), Word (2x).
Table 3: Phase II.1

Watch Time

Afternoon

Noon - 4:00 p.m.

Dog

4:00 p.m. - 8:00 p.m.

First

8:00 p.m. - Midnight

Middle

Midnight - 4:00 a.m.

Morning

4:00 a.m. - 8:00 a.m.

Forenoon

8:00 a.m. - Noon

1 Coding unit: seaman's watch (3x).

226

Table 4: Phase II Code.'

A. Concentration

Ice Free

Open Water

Very Open Ice

Open Ice

Close Ice

Very Close Ice

Consolidated/Compact Ice

Floe Size

Giant Floe

Vast Floe

Big Floe

Medium Floe

Small Floe

Ice Cake

Small Ice Cake

Openings

Crack

Open Lead

Blind Lead

Shore Lead

Flaw Lead

Arrangement

Ice Field

Belt

Tongue

Strip

Ice Edge (compacted)

Ice Edge (diffuse)

Concentration Boundary

Motion

Diverging

Converging

Shearing

No compulsory codes. Coding units: seaman's watch (3x).

227

Manual) for the terms. Each coder applied this system to all of the watches on three separate
occasions so that their consistency could be determined.

The analysis of these sessions followed a more complex process that eliminated biases inherent
in the use of percent agreements. This was appropriate here because the development and
application of the categories was more structured than in Phase I. In this case, Krippendorff s
agreement coefficient was calculated by using the following equations.

a=l-_

Where:

a
D

= agreement coefficient
= observed disagreements

= expected disagreements
and

o jLu I—u v

Where: x bc = number of disagreements in a matrix of
category codes (or coders)

- total of the marginal entries
and

Where: x5 x>c

x
m

D.=XZ

x, x

b. .c

x (x -m+1)

b c

= the products of all possible marginal
entries of the matrix

= the marginal total
= number of category codes (or coders)

The resulting coefficient is a number between 0 and 1 which, when multiplied by 100, gives the
percentage by which the agreements are better than chance. Therefore, when the coefficient is
0, then any agreement is completely by chance. When the coefficient is 1, the agreements are
based entirely on the coders' judgements with no degree of chance.

When this was applied to the intercoder agreements (for category A - Concentration), the
coefficient was 0.468 or 47% better than chance. The average intracoder agreement was 0.591.
It was then decided that these figures could be improved by modifying the categories since the
problem was not due to unskilled coders. One of the many advantages of this agreement
coefficient is that it can also be used as a diagnostic device in the restructuring of the categories
(or reselection of coders, if necessary).

228

When the cause of low coefficient values is due to a problem with the categories, it is usually
because the distinctions between the codes are not sufficient. This can be remedied by combining
those codes that are most frequently confused with one another. Figures 6a-d provide an example
of this testing procedure. Figure 6a is the basic matrix for the seven codes in the concentration
category. The numbers in the cells indicate the frequencies with which each was used, so that all
of the diagonal entries are the numbers of agreements among all five coders for each code, and
the off-diagonals are the disagreements. The coefficient for this matrix was calculated to be
0.468. Figure 6b shows the matrix if it was collapsed into two codes: no ice (1) and ice (2-7).
Intuitively, this would be expected to substantially increase the agreement coefficient. However,
because there is no longer a bias in favour of fewer categories the value was increased by only
0.086. Figure 6c shows another regrouping into three codes: no ice (1), general ice descriptions
(2-6), and complete ice coverage (7). This raised the coefficient by only 0.016. Finally, the codes
were regrouped (Figure 6d) into four codes: no ice (1), open ice (2-3), close ice (4-5), and
consolidated ice (6-7) and this increased the coefficient by 0.203 so the value became 0.689
(almost 70% better than chance). It should be stressed here that the coefficients in Figures 6b,
c, and d were all calculated from the original matrix and not by recoding. This process was also
applied to categories B (floe size) and D (arrangement) with increases in the coefficient of 0.333
and 0.479 respectively.

Phase III

The coding system for this phase resulted directly from the regroupings discussed above and is
presented in Table 5. Category C (openings) was omitted due to infrequent usage by the coders,
and the other four categories were regrouped as illustrated by comparing Tables 4 and 5. This
coding session was only repeated twice because the coders' consistencies had been sufficiently
tested by this point. The same definitions and diagrams were used here as in Phase II, the major
difference being that category A (concentration) was compulsory. That is, a code designation was
required for this category for every watch of every day. The agreements were analyzed as in
Phase n. The coefficients for category A are given in Table 6. Because the other three categories
were so rarely used, agreement coefficients were calculated only for category A.

Two observations are clear from Table 6. First, regrouping raised the agreement level by 21%.
Secondly, although the averages of the coefficients for Phase III were lower than for the Phase
II regrouped figures, they differed by only 2%. Therefore it is possible to use the calculated
regroupings as a prediction for the Phase III coding agreements, and the third phase of coding
could actually be eliminated.

It was concluded that the Phase III coding system produced an acceptable level of reliability since
an average of only 37% of the agreements were made by chance and the remaining 63% were
reliable agreements. As a result, the seaman's watch and the concentration category were adopted
as the basis for the sea-ice reconstruction.

229

a COINCIDENCE MATRIX

PHASE 2 : CATEGORY A - CONCENTRATION

b COINCIDENCE MATRIX

PHASE 2 : CATEGORY A - CONCENTRATION

Agreement Coefficient = .554
Categories : No Ice & Ice

168

105

275

105

138

339

434

570

218

107

404

107

208

370

2068

[l68i

105

275

105

$138

339

l86

434

570

I 10

218

107

404

1 0

107

208

370

1°

1 1

2068

e COINCIDENCE MATRIX

PHASE 2 : CATEGORY A - CONCENTRATION
Agreement Coefficient = .484

Categories : No Ice. General Ice, Consolidated Ice

d COINCIDENCE MATRIX

PHASE 2 : CATEGORY A - CONCENTRATION
Agreement Coefficient = .689

Categories : No Ice, Open Ice, Close Ice. Consolidated

1168!

105

275

105

1 1 38

339

I 86

434

570

218

107

404

107

208

370

—

16
aaasa

1 1

1 I

2068

1168 1

105

275

105

1 1 38

86 |

339

434|

570

218

107

404

107

208

370

\ 12

1 1

o 1

2068

Figure 6: Coincidence matrices and agreement coefficients, (a) Basic matrix; (b) Two-category matrix;
(c) Three-category matrix; (d) Four-category matrix.

230

Table 5: Phase III Code.1

A. Concentration

1. Ice Free

2. Open Water/Very Open Ice

3. Open Ice/Close Ice

4. Very Close/Consolidated/Compact Ice

B. Floe Size

1. Small Ice Cake

2. Ice Cake

3. Medium/Small Floe

4. Big Floe

C. Arrangement

1 . Strip/Diffuse Ice Edge/Concentration Boundary

2. Belt

3. Tongue

4. Ice Field/Compacted Ice Edge

D. Motion

1 . Diverging

2. Compacting

1 Category A compulsory. Coding units: seaman's watch (2x).

Table 6: Intercoder Agreement Coefficients. Category A - Concentration.

Coding Session

Phase II Regrouped Phase II Phase III

.422 .653 .666

.463 .653 .603

.468 .689 1

▼ ▼

Average Differences

+ .214 -.019

Phase III was only repeated twice.

231

Concluding Remarks

Although this case study was not directed specifically to the climatic anomaly of 1816, its
relevance pertains to the entire time frame in which this event occurred. There is a potentially
large volume of climatic information relating to key volcanic episodes that is in a descriptive
format. This study provides a solution to the problem of interpreting this type of information
reliably, an approach that in the past has been superficially addressed. Reconstructions based on
categories that are developed from thoroughly-tested evolutionary process provide information
that describes the reliability with which the original documents were interpreted. Furthermore,
the results of these reliability tests must accompany the reconstruction so that the question of an
acceptable level of reliability is relegated to the user of the reconstruction to a certain degree.
This does not mean that any agreement coefficient should be accepted by the researcher. In the
search for an acceptable level of reliability, an attempt should be made to balance the amount of
information obtained against the degree of objectivity by which it was derived. In this study, a
considerable amount of information was discarded throughout the testing procedure from the first
phase to the final code so that sea-ice concentration was the only category to be used in the final
reconstruction.

232

River Ice and Sea Ice in the Hudson Bay Region during the Second
Decade of the Nineteenth Century

A.J.W. Catchpole1
Abstract

Analysis of documentary sources in the Hudson's Bay Company Archives has provided records
of river- and sea-ice conditions in the Hudson Bay region during the eighteenth and nineteenth
centuries. These include six records of dates of first-breaking and first-freezing of routes to the
bayside trading posts. Several different validity tests have been applied to these data, and the
results of these tests generally indicate the data are valid measures. The values of river- and sea-
ice data in each year from 1810 through 1820 are compared with their values during the whole
period of record. This enables the identification of years with anomalously early or late dates of
breaking and freezing, and years with severe summer sea ice. Evidently, exceptionally cold
summer weather occurred in the second decade of the nineteenth century. This was not initiated
after the eruption of Tambora in April 1815, but was first apparent in 1811 and 1812. However,
the most severe summer cold in the decade occurred in the two years following the eruption.

Introduction

In subarctic regions the dispersal of ice in spring and freezing of water bodies in fall are
intimately linked to weather and climatic conditions. Anomalous weather in a particular year may
cause exceptionally early or late breaking and freezing of rivers, lakes and seas. Ice observations
therefore figure prominently among the routine climatologic and oceanographic observations made
by the nations fringing the poles. Another property of ice is that it occurs in forms vividly
apparent to casual observers, and under circumstances where it can severely restrict their physical
activities. For these reasons, informal descriptions of ice also occur prominently in written
historical sources that contribute to the reconstruction of climates in the recent past. So it was that
the daily journals kept by servants of the Hudson's Bay Company graphically described the long
anticipated breaking of the rivers in spring and their equally vital refreezing in fall. Likewise, the
log-books of the Company's supply ships that annually sailed the ice-congested waters of Hudson
Strait and Hudson Bay gave frequent, detailed descriptions of the ice which imperilled their
passage.

The Ice Records

These sources have yielded six records of dates of first-breaking and first-freezing of the estuaries
of rivers draining into Hudson Bay and three records of summer sea-ice severity encountered
along portions of the sailing routes to the bayside trading posts. The various records commenced
in the early or mid-eighteenth century, and in most cases ended in the latter part of the nineteenth
century. Table 1 lists for each record its location, the year when the record commenced and
ended, the total number of years in which ice data have been reconstructed and the source in
which the reconstruction was originally published. Three stages in the seasonal development of
river ice are dated:

Department of Geography, The University of Manitoba, Winnipeg, Manitoba R3T 2N2, Canada.

233

1. date of first-breaking - the first day on which any evidence of breaking was observed,
irrespective of whether or not the river remained broken thereafter;

2. date of first partial freezing - the first day on which the river was observed to become
partially frozen, irrespective of the spatial extent of the ice cover or its continuity thereafter;

3. date of first complete freezing - the first day on which the entire surface of the river was
frozen, irrespective of whether or not it remained completely frozen thereafter.

Each of these historical records was derived from daily journals written at trading posts located
in the estuaries of rivers draining into Hudson Bay (Figure 1). In several of these estuaries the
locations of the posts changed from time to time, but the records derived in the Severn, Albany,
Moose and Eastmain estuaries are each based on single post locations. The Churchill journals
were written both at the Old Fort, located inside the estuary, and at Fort Prince of Wales situated
on an exposed promontory where the north shore of the estuary protrudes into Hudson Bay. The
Churchill first-freezing data used in this paper were those reconstructed from the Old Fort
journal, and the first-breaking data were those derived at Fort Prince of Wales. In the Hayes
River estuary the location of York Factory was changed in 1791 to a site very close to the former
on the north shore of the estuary.

Table 1: Historical Records of River Ice and Sea Ice Derived from Hudson's Bay Company Archives.

Dates of First-Breaking of River Estuaries

Number

Location

Limits

of Years

River

of Record

Sources

Churchill

Churchill Old Fort

1720-1866

110

Moodie and Catchpole (1975)

Fort Prince of Wales

1731-1861

107

Hayes

York Factory 1

1715-1790

Moodie and Catchpole (1975)

York Factory 2

1791-1851

Moodie and Catchpole (1975)

Severn

Fort Severn

1763-1939

104

Magne (1981)

Albany

Fort Albany

1722-1939

190

1722-1866 (Moodie & Catchpole 1975);

1872-1939 (Magne 1981)

Moose

Moose Factory

1736-1871

133

Moodie and Catchpole (1975)

Eastmain

Eastmain House

1743-1939

109

Magne (1981)

234

Table 1: (cont'd)

Dates of First Partial Freezing of River Estuaries

River

of Record

T imitc

of Record

Number

r\T ir pore

KJl I Cell S

of Record

Sources

Churchill

Churchill Old Fort

Port Prinrp of ^VaIps

1718-1866
1731-1845

69
42

Moodie and Catchpole (1975)

Moodip and Catrhnnlp MQ7S^

Hayes

York Factory 1
York Factory 2

1714-1790
1791-1850

73
44

Moodie and Catchpole (1975)
Moodie and Catchpole (1975)

Severn

Fort Severn

1761-1940

100

Magne (1981)

Albany

Fort Albany

1721-1938

180

1721-1867 (Moodie & Catchpole 1975);
1872-1938 (Magne 1981)

Moose

Moose Factory

1736-1870

132

Moodie and Catchpole (1975)

Eastmain

Eastmain House

1743-1940

Magne (1981)

Dates of First Complete

Freezing of River Estuaries

River

Location
of Record

Limits
of Record

Number
of Years
of Record

Sources

Churchill

Churchill Old Fort
Fort Prince of Wales

1718-1865
1722-1852

95
55

Moodie and Catchpole (1975)
Moodie and Catchpole (1975)

Hayes

York Factory 1
York Factory 2

1714-1792
1793-1851

76
39

Moodie and Catchpole (1975)
Moodie and Catchpole (1975)

Severn

Fort Severn

1760-1940

Magne (1981)

Albany

Fort Albany

1721-1921

178

1721-1864 (Moodie & Catchpole 1975);
1872-1921 (Magne 1981)

Moose

Moose Factory

1739-1861

117

Moodie and Catchpole (1975)

Eastmain

Eastmain House

1743-1940

Magne (1981)

235

Table 1: (cont'd)

Summer Sea-Ice Severity Indices

Region

Number
Limits of Years
of Record of Record

Sources

Hudson Strait

Eastern Hudson Bay
Western Hudson Bay

1751-1889 137

1751-1870 108
1751-1869 111

175 1-1 870 (Catchpole and Faurer 1983);
1 87 1 - 1 889 (Catchpole and Hanuta 1 989)

Catchpole and Halpin (1987)

Catchpole and Hanuta (1989)

90°W 80° 70° 60°W

Figure 1: Location map showing sailing routes through Hudson Strait, across eastern Hudson Bay to
Moose, and across western Hudson Bay to York and Churchill.

236

The three sea-ice severity records refer not to point locations hut to the three portions of the
sailing-ship route (Figure 1). These records were reconstructed from descriptions of ice given in
the supply ships' log-books. These ice-severity indices are numerical in form but they function
as ordinal not interval data. As such, the indices rank the years on the basis of summer-ice
severity, but they are not numerical measures of the quantities of ice present in each summer.
The frequency distributions of the ice indices are highly skewed, with very high proportions of
small values and a few very large values. This property implies that the indices discriminate more
accurately between the ranking of the few severe ice years than between that of the large number
of moderate and light ice years.

Quality of Ice Records

The objective of this paper is to use the records listed in Table 1 to determine whether the river-
and sea-ice conditions in the second decade of the nineteenth century were in any respects
anomalous when compared with the ice conditions observed throughout the periods of record. In
view of this objective it is pertinent to comment briefly on the quality of these historical data.
Two aspects of the quality of climatic data derived from historical sources should be considered.
These are the reliability of the method of derivation and the validity of the data derived. The
reliability of the method determines the degree to which similar results will be obtained when the
same method is applied to the same sources by the same person, or by different people with
similar training. The validity of the data determines the degree to which the results are true
measures of what they are intended to measure. There has been no fully comprehensive testing
of the quality of these historical river- and sea-ice data. However, several studies have yielded
information that bears upon their reliability and validity.

The derivation of the breaking and freezing dates of the river estuaries (Moodie and Catchpole
1975) included reliability testing as one of its major aspects. The test results showed that high
degrees of reliability were obtained when dates based on direct dating categories were derived
for places where the journals were kept. Much lower levels of reliability were obtained for dates
based on less direct information. Marcia Faurer (this volume) is developing and applying an
innovative approach to testing the reliability with which sea-ice data can be derived from sailing
ships' log-books.

The validity of river-ice dates has been tested internally by examining the spatial homogeneity
between similar dates derived at adjacent estuaries. These tests found high correlations between
the dates of first-breaking at Fort Albany and Moose Factory and supported the conclusion that
these are, therefore, true measures of the actual breaking dates in these river estuaries (Moodie
and Catchpole 1976). Some studies have compared selected river-ice dates with tree-ring data
derived from trees growing in the vicinity of the river estuaries. These studies were not designed
as tests of the validity of the ice data but they do detect similarities between the trends revealed
by tree-ring and ice data. In so doing they provide rudimentary indications of the validity of the
ice data tested against external criteria. This approach is exemplified by a study of ice conditions
in the Churchill River estuary conducted by Jacoby and Ulan (1982). This study used tree-ring
data from near Churchill. It found a multiple correlation coefficient of 0.69 between tree growth
and the date of complete freezing at Fort Prince of Wales during 1741-64. Jacoby and Ulan
(1982) used this relationship to derive dates of complete freezing from tree-ring data in the period
1680-1977. In his reconstruction of temperatures in the Hudson Bay region during the past three
centuries, Guiot (1986; this volume) assembled a database including early instrumental
temperature observations, tree-ring data and river-ice dates.

237

Significant correlations were found between several of the records of the first-breaking and
freezing of river estuaries and other records in this database (Guiot 1986, pp. 13, 19). Dates of
first partial freezing and first complete freezing were generally found to be positively correlated
with autumn temperatures measured at York and Churchill, whereas dates of first-breaking were
generally negatively correlated with spring temperatures. Some of the tree-growth records were
negatively correlated with the date of first complete freezing, and this finding is consistent with
the results obtained by Jacoby and Ulan (1982). Lough and Fritts (1987; Lough this volume) used
North American tree-ring data to assess the possible effects of volcanic eruptions on North
American climate during 1602-1900. In this study they employed the mean dates of first-breaking
and first complete freezing of the James Bay estuaries as "independent temperature records
outside the area covered by the arid site tree-ring reconstructions." Using superposed epoch
analysis, Lough and Fritts detected changes in ice dates following major volcanic eruptions that
were consistent with the observed changes in tree growth.

Wilson (1988; this volume) has derived summer thermal indices for the southeast coast of Hudson
Bay in the nineteenth century, using a miscellany of historical evidence in the Hudson's Bay
Company Archives. A preliminary study of these indices shows that they may afford an indirect
means of testing the validity of sea-ice data, in so far as anomalous summer cold in this region
may be a result, or a cause, of severe summer ice on adjacent seas. This study involved a
comparison between the incidence of severe ice years and negative anomalies in the thermal
indices for May to June (Figure 2 A) and May to October (Figure 2B). The May to June data
were selected for this comparison because the ships' log-books were not among the historical
sources used to derive these indices. The May to October data were selected because Wilson
(1988, p. 13) considered that the index is most accurate over the entire summer season. However,
the sea-ice indices and May to October thermal indices are not entirely independent because the
ships' log-books did play a minor role as sources in the derivation of the mid-summer thermal
conditions.

Figure 2 comprises graphs of Wilson's thermal indices upon which are superimposed vertical bars
identifying severe ice years. In this context a severe ice year is defined as one of the years having
the 10 highest ice indices in each of the three ice records derived for Hudson Strait, eastern
Hudson Bay and western Hudson Bay. The 10 highest indices are based on the entire period of
the sea-ice records, not the period 1800-70. A vertical bar on Figure 2 indicates that severe ice
occurred in that year in Hudson Strait or in eastern or western Hudson Bay. It is judged to be
appropriate to consider these three records together in this way and not separately. Severe ice in
Hudson Strait could retard the entry of ships into the bay to such a degree that they would not
encounter the bay ice in July and August, but rather in September. At this time even severe late
summer ice is generally cleared from the bay. Furthermore, the eastern and western parts of the
bay are not separate entities in the context of ice clearing, but rather the lateral limits of the
waters in which the last remnants of ice tend to congregate under the influence of prevailing
winds and currents (Danielson 1971). In years with zonal atmospheric circulation these remnants
tend to be driven towards the east and accumulate in the sailing route to James Bay. A meridional
atmospheric circulation permits late ice to remain in the west in the path of ships sailing to
Churchill or York Factory.

Figure 2 reveals a tendency for severe ice years to concur with periods having negative thermal
indices. This is most apparent in the middle of the second decade of the century, in the late 1830s
and in the early to mid- 1840s. It is noteworthy that these are generally periods in which Wilson

238

(1988, pp. 7, 8) noted the quality of the thermal indices as good to excellent. It is not appropriate
to numerically evaluate the correlation between these data because the ice indices are ordinal not
interval data.

A. MAY-JUNE THERMAL INDICES

6-i

4 -
2-

-2-
-4-
-6-

■8-

1800

1810

1820

1830

1840

1850

1860

1870

B. MAY-OCTOBER THERMAL INDICES

4 -
2-

-4-

1800 1810 1820

QUALITY OF THERMAL DATA

1830

excellent
good

intermediate

least

1840 1850 1860 1870

INCIDENCE OF SEVERE ICE INDICES

year in which one severe ice index
occurred

year in which two severe ice indices
occurred

Figure 2: Thermal indices for the southeastern coast of Hudson Bay (from Wilson 1988), and years with
severe summer ice in Hudson Strait and Hudson Bay, 1800-70. The thermal indices are
estimates of departures from the 1941-70 normals of temperatures in the May to June (A) and
May to October (B) periods. The quality of these indices in different time intervals was
assessed by Wilson (1988, p. 7-8). A severe ice year is defined as a year having one of the 10
highest ice indices, in the period 1751 to 1870, within each of the three ice records.

239

Ice Conditions, 1810-20

River- and sea-ice conditions in each year from 1810 through 1820 are evaluated in Tables 2A-C
and 3. The data given in these tables compare the ice condition in each year with the range of
values of that condition reconstructed over the whole period of record. In the case of the river-
ice dates (Tables 2A-C) the comparison is made by the calculation of the parameter Z.1 This
enables the identification of years with anomalously early or late dates of breaking and freezing.
Table 4 lists these years and distinguishes between anomalies having less than 1, 2.5 and 5%
probabilities of occurring by chance. Table 3 gives the rank order of occurrence of each sea-ice
severity index among the indices reconstructed for the whole period of record. Table 4 identifies
the years in which the sea-ice index was ranked among the upper 10 values in each record.

Fourteen of the river-ice records are designated anomalous in Table 4, and all of these are
indicative of summer cold with significantly late-breaking and early-freezing. During this decade
there was no occasion of early-breaking or late-freezing that produced a Z value so large that
there was only a 5% probability of its occurring by chance. The greatest anomalies, in frequency
and amount, were those of retarded first-breaking in 1817 and 1812. In 1817 the date of first-
breaking was anomalously late in all of the river estuaries from which a date could be obtained
in that year. The record was interrupted in 1817 at York and Severn (Table 2A). In 1812 this
date was anomalously late in four estuaries but not at Moose or Eastmain. Furthermore, at York
and Severn, the 1812 anomalies exhibited 5% probabilities of occurrence by chance, whereas all
of the 1817 anomalies exceeded this level of significance. First partial freezing was significantly
early at Eastmain in 1817 and at Churchill in 1811. First complete freezing was early at York
and Eastmain in 1811 and at York in 1817.

This decade was marked by severe late-summer ice in eastern Hudson Bay in 1813 and by a
cluster of high sea-ice indices in 1815 to 1817. This cluster included the highest ice index derived
in Hudson Strait (1816), as well as severe ice in western Hudson Bay in 1815 and in eastern
Hudson Bay in 1816 and 1817. 1816 provides a case in which the passage of the ships through
Hudson Strait was so greatly delayed that they apparently entered the bay so late as to reduce
their ability to monitor a mass of ice that persisted late in the summer within the bay. This ice
was located in the east across the sailing route to James Bay. The ship in question (the Emerald)
rounded Mansell Island and entered Hudson Bay on 7 September. This was 25 days later
(standard deviation 9.7) than the mean date on which ships bound for Moose Factory entered the
bay in the period 1751-1870. During this delayed passage to Moose in 1816, the Emerald
encountered ice which yielded the seventh largest index (Table 3). Probably the 1816 ice in
eastern Hudson Bay would have ranked even higher if the Emerald had sailed these waters closer
to the normal sailing date. In 1816, the Prince of Wales sailed to York Factory. This ship also
entered the bay on 7 September. However, it encountered no ice on its passage to the west coast
and there is, therefore, no evidence that this ship would have encountered exceptionally late ice
if it had sailed into the bay earlier than this late date.

1 Z = xjx
a

where:

x = date of breaking (first partial freezing, first complete freezing) in a particular year;
^ = mean date for whole period of record;

a = standard deviation from this mean for whole period of record.

240

Table 2A: Dates of First-Breaking of River Estuaries, Standard Units Z.1

Estuary (n = number of years of record)

Churchill2

York3

Severn

Albany

Moose

Eastman

n=107

n = 45

n=104

n=190

n=133

n=10<

1810

+ 0.31

+ 0.05

+ 0.71

-1.49

-1.61

1811

+ 0.72

+ 0.37

-0.28

-0.39

-0.79

1812

+ 2.67

+ 1.98

+ 2.17

+ 2.38

+ 1.49

+ 1.76

1813

+ 0.45

-1.12

-0.72

-0.75

1814

+ 1.27

+ 1.07

-0.01

1815

+ 1.84

+ 0.60

+ 1.44

+ 1.60

+ 2.05

+ 1.53

1816

-1.87

+ 0.72

+ 0.83

+ 1.41

1817

+ 2.40

+ 2.27

+ 2.30

+ 3.03

1818

-0.95

+ 0.72

-0.39

-0.21

1819

+ 0.58

-0.91

-1.48

+ 0.06

+ 0.10

-0.44

1820

-0.81

-1.12

-1.69

-1.49

-1.13

1 Z=xji-

2 Estuary of Churchill River at Fort Prince of Wales.

3 Estuary of Hayes River at York Factory 2.

Table 2B: Dates of First Partial Freezing of River Estuaries, Standard Units Z.1

Estuary (n = number of years of record)

Churchill2 York3 Albany Moose Eastmain

n = 69 n = 44 n=180 n=132 n = 98

1810

-0.77

-0.12

1811

-2.20

-1.29

-1.32

-1.12

-1.67

1812

-0.18

-1.15

+ 0.16

1813

-0.94

-1.07

-1.35

-0.83

1814

+ 0.96

-1.73

+ 0.39

+ 0.63

+ 1.31

1815

+ 1.29

+ 0.02

-0.30

-0.24

1816

+ 1.37

+ 1.19

1817

-0.43

-0.58

-0.88

-2.26

1818

-0.43

+ 0.71

+ 1.24

+ 0.86

+ 1.67

1819

-0.30

+ 1.15

-0.46

-0.77

-0.48

1820

-0.14

+ 0.15

-0.42

-0.48

1 Z=x-u-
a

1 Estuary of Churchill River at the Old Fort.
3 Estuary of Hayes River at York Factory 2.

241

Table 2C: Dates of First Complete Freezing of River Estuaries, Standard Units Z.1

Estuary (n = number of years of record)

Churchill2
n = 95

York3
n = 39

Albany
n=178

Moose
n=117

Eastmair
n = 97

1810

1811

-2.20

-1.94

-1.82

-2.15

1812

-0.94

-2.13

-0.83

-1.15

1813

-0.68

1814

+ 0.60

+ 0.46

+ 1.02

+ 0.84

1815

-0.47

-0.83

-1.49

-0.52

1816

-0.24

+ 0.56

1817

-1.20

-1.70

1818

+ 0.74

+ 1.68

+ 1.30

+ 1.25

1819

-1.11

-1.34

1820

-0.83

-0.16

-0.34

1 Z=x-u.

2 Estuary of Churchill River at the Old Fort.

3 Estuary of Hayes River at York Factory 2.

Table 3: Summer Sea-Ice Severity Indices, Annual Ranking.

Location (n = number of years of record)

Eastern Western
Hudson Strait Hudson Bay Hudson Bay

n=137 n=108 n=lll

1810

1811

1812

1813

101

1814

101

1815

1816

101

1817

1818

1819

101

1820

135

242

Table 4: Incidence of Anomalous River-Ice Dates and Severe Sea Ice During 1810-20. The Probabilities
of River-Ice Anomalies are Based on the Standard Units Z (Tables 2A-C). The sea-ice anomalies
are the years having one of the 10 highest ice-severity indices in each of the three records. The
fractions given compare the rank with the number of years in the record.

1810

1811

1812

1813

1814

1815

1816

1817

1818

1819

1820

y) i First
w | Breaking

3 1

uj |-

> First
oc | Partial
^ | Freezing

£ L

UJ |

£ | First
_ 1 oompieie
a j Freezing

EARLY AT
Churchill

EARLY AT-

York
Eastmain

LATE AT

CHURCHILL

ALBANY

York
Severn

EARLY AT
York

LATE AT
Moose

LATE AT

EASTMAIN

CHURCHILL
ALBANY
MOOSE

EARLY AT
EASTMAIN

J> 1 Hudson .
uji- 1 Strait

COqc i

K > 1 Hudson Bay
puj ]_ (East)

UJ 1

inO Hudson Bay
- | (West)

SEVERE
2/108

SEVERE:
6/111

SEVERE:
1/137

SEVERE:
7/108

SEVERE
8/108

PROBABILITY THAT THIS ANOMALY OCCURRED BY CHANCE
CHURCHILL : Less than 1% ALBANY : Less than 2.5% York : Less than 5%

Conclusions

The river- and sea-ice data presented here indicate that in the second decade of the nineteenth
century cold summer weather was not initiated after the eruption of Tambora in April 1815, but
was first apparent in 1811 and 1812. However, this evidence does show that the most severe
summer cold in that decade occurred in the two years following the eruption.

The first of these cold episodes commenced in 1811 with early first partial freezing at Churchill
and early first complete freezing at York and Eastmain. This was followed in the spring of 1812
with late first-breaking at Churchill and Albany and with late breaking, though less delayed, at
York and Severn. In the fall of 1812 first complete freezing occurred early at York.

An isolated case of severe sea-ice occurred in eastern Hudson Bay in 1813, and this was followed
by a cluster of years with severe ice in 1815 (western Hudson Bay), 1816 (Hudson Strait and
eastern Hudson Bay) and 1817 (eastern Hudson Bay). This period culminated in late first-
breaking in 1817 at Eastmain, Churchill, Albany and Moose. In the fall of 1817 early first partial
freezing occurred at Eastmain. There were gaps in the historical record during both of these cold-
summer periods, and these were most prominent in 1816 and 1817. In particular, data on first
partial freezing and first complete freezing are unavailable for Churchill and York in 1816, and
no river-ice data are available for York in 1817.

243

References

Catchpole, A.J.W. and M.A. Faurer. 1983. Summer sea-ice severity in Hudson Strait,
1751-1870. Climatic Change 5:115-139.

Catchpole, A.J.W. and J. Halpin. 1987. Measuring summer sea-ice severity in eastern Hudson
Bay 1751-1870. Canadian Geographer 31:233-244.

Catchpole, A.J.W. and I. Hanuta. 1989. Severe summer ice in Hudson Strait and Hudson Bay
following major volcanic eruptions, 1751 to 1889 A.D. Climatic Change 14:61-79.

Danielson, E.W. 1971. Hudson Bay ice conditions. Arctic 24:90-107.

Guiot, J. 1986. Reconstruction of temperature and pressure for the Hudson Bay Region from
1700 to the present. Canadian Climate Centre Report No. 86-11:1-106.

Jacoby, G.C. and L.D. Ulan. 1982. Reconstruction of past ice conditions in a Hudson Bay
estuary using tree rings. Nature 298:637-639.

Lough, J.M. and H.C. Fritts. 1987. An assessment of the possible effects of volcanic eruptions
on North American climate using tree-ring data, 1602 to 1900 A.D. Climatic Change
10:219-239.

Magne, M.A. 1981. Two centuries of river ice dates in Hudson Bay region from historical
sources. MA. thesis, University of Manitoba, Winnipeg. 78 pp.

Moodie, D.W. and A.J.W. Catchpole. 1975. Environmental data from historical documents by
content analysis: freeze-up and break-up of estuaries on Hudson Bay 1714-1871. Manitoba
Geographical Studies 5:1-119.

. 1976. Valid climatological data from historical sources by content analysis. Science

193:51-53.

Wilson, C.V. 1988. The summer season along the east coast of Hudson Bay during the nineteenth
century. Part III. Summer thermal and wetness indices. B. The indices, 1800 to 1900.
Canadian Climate Centre Report No. 88-3:1-42.

244

The Climate of the Labrador Sea in the Spring and Summer of 1816,
and Comparisons with Modern Analogues

John P. Newell1

Abstract

The wide range of natural variability in climatic conditions at the local and regional scales makes
it necessary to examine data from as large an area as possible in order to determine the
significance of past departures from present-day conditions. Many authors have demonstrated that
the spring and summer of 1816 were among the coldest ever recorded in the region extending
from the northeastern United States to Hudson Bay. Recent research on tree rings indicates that
climate may not have been as severe in the western United States and Canada. This study
examines proxy-climatic data for northeastern North America, extending from southeastern
Newfoundland to Hudson Strait and including the waters of the Labrador Sea, in an effort to
develop a more continental view of climate during this critical period.

The sources investigated include: weather narratives from both Newfoundland and Labrador; a
daily weather diary from eastern Newfoundland; and sea-ice records for the waters adjacent to
Newfoundland and Labrador. The study demonstrates that, during the spring and summer of
1816, climatic and sea-ice conditions in northern Labrador were among the most severe ever
recorded; however, farther south in Newfoundland, conditions were by no means as severe, and
may have been near nineteenth century normals.

The 1816 patterns of climatic and sea-ice conditions in Newfoundland and Labrador are compared
with recent (post- 1950) patterns of temperature, precipitation and sea-ice conditions in eastern
North America to determine if modern analogues exist. This comparison indicates that conditions
in 1816 have no clear analogues in the recent climatic record. However, there are patterns that,
while not as severe, do provide some indications of the nature of the circulation in 1816. These
patterns indicate that the circulation during the summer of 1816 was similar to the present
normals for March and April. This agrees with the July circulation pattern for 1816 presented
by Lamb and Johnson (1966).

Introduction

The unusual character of the summer of 1816 in the Labrador Sea is demonstrated by the
following report from the records of the Moravian Church which operated several missions along
the Labrador coast: "The Jemima [the moravian mission ship] arrived in the river [Thames] from
Labrador, after one of the most dangerous and fatiguing passages ever known. As in almost every
part of Europe, so in Labrador, the elements seem to have undergone some revolution during the
course of last summer" (Periodical Accounts, Vol. VI, p. 263).

Modern research has demonstrated that the summer of 1816 was unusually cold in Europe
(Manley 1974; Kelly et al. 1984; Briffa et al. 1988), eastern United States (Stommel and
Stommel 1979; Ludlum 1966); and Hudson Bay (Wilson 1983; Catchpole 1985). Other authors

34 Cornwall Crescent, St. John's, Newfoundland A1E 1Z5, Canada.

245

have demonstrated that sea-ice conditions in both Hudson Strait (Catchpole and Faurer 1985) and
the Labrador coast (Newell 1983) were extremely severe during the summer of 1816. By
comparison, sea-ice conditions in the East Greenland Sea (Scoresby 1820) and near Iceland
(Lamb 1977; Ogilvie, this volume), while more severe than normal, did not reach the record
conditions experienced in eastern North America.

The only previous study (Lamb and Johnson 1966; Lamb, this volume) that directly considers
climatic conditions in the Labrador Sea during 1816 is an analysis of January and July global sea-
level pressure patterns for the years 1750 to 1962. It includes a map of July 1816 circulation over
the North Atlantic indicating that a 1002 mb low-pressure centre was situated over the Labrador
Sea, giving a northerly flow along the Labrador coast. This circulation pattern is more
representative of conditions in April than of the normal circulation in July. It should be noted that
Lamb and Johnson provide maps showing that the Labrador Sea was outside the limits of reliable
isobars until the 1870s. While the exact data used to construct their map for 1816 are not given
in the report, other sources (Lamb and Johnson 1959, 1961) indicate that it was likely based on
wind data from New England and possibly Greenland.

This paper presents the results of an analysis of proxy-climatic records from areas surrounding
the Labrador Sea (Newfoundland, Labrador, Hudson Strait and southwestern Greenland) and an
attempt to reconstruct the atmospheric circulation pattern in this region for June 1816. In
addition, temperature patterns over the area in June 1816 are used to select modern analogues for
the 1816 circulation pattern. These modern analogues are then compared with the reconstructed
circulation pattern. The study area and locations noted in the text are shown in Figure 1.

Analysis of Historical Data

The following brief review of the history of the study area in 1816 provides an indication of types
of proxy-climatic data sources available. At the start of 1816, Newfoundland was in the midst
of a financial crisis caused by the fall in fish prices after the end of the War of 1812, and in
February 1816 a major fire struck St. John's, the capital of the island. At this time the main
economic activity in Newfoundland was the inshore cod fishery. The only other significant
economic activity was the seal "fishery" carried out off the northeastern coast each spring.
Farther north along the Labrador coast, the Moravian Church operated missions it had established
during the late eighteenth century. These missions were supplied each spring by a mission ship
that sailed directly to Labrador from England. At the same time the Moravians also operated a
number of missions in southwestern Greenland that were supplied by Danish ships sailing from
Denmark to the Greenland settlements. Whaling ships from Britain also operated off the west
coast of Greenland each spring. The only other significant shipping activity in the study area at
this time involved Hudson's Bay Company ships that sailed from England to Hudson Bay each
spring and returned in the fall.

A review of material available in the Newfoundland Archives revealed that government
correspondence from this period is rather limited. This is partly due to the fact that prior to 1818
the governor was only resident in Newfoundland during the summer. The only pertinent remark
was: "The weather during the greater part of the season [summer 1816] has been particularly
unfavourable for the curing [the cod was dried in the sun] of fish" (Report of Fishery, December
1816, Government Letter Book, Newfoundland Archives). This situation could result from either
damp weather or calm weather with clear skies. Analysis of historical catch statistics for cod in
Newfoundland waters (Forsey and Lear 1987) indicate that 1816 was a relatively good year.

246

Figure 1: Study area.

247

While no statistics on the seal catch are available for 1816, available data do not point to a bad
year. The records indicate however a very low catch in 1817, which was attributed to severe ice
conditions.

A weather diary kept at Trinity, Newfoundland by the firm of Slade and Kelson provides the best
information on the climate of Newfoundland in 1816. A review of the weather remarks and
rain/snow frequencies given in the diary do not provide any evidence for cold conditions during
the spring or summer of 1816. Analysis of the daily reports of wind for June 1816 indicate a high
frequency from the southwest (55%) compared to present day normals for Bonavista,
Newfoundland (less than 30%) and compared to Trinity in 1817 (48%) and 1818 (36%). A
comparison of air temperature versus wind direction for St. John's, based on modern data,
indicates that the two parameters are closely linked (Figure 2). Southwest winds are clearly warm
winds, so it was likely that southeastern Newfoundland experienced normal to above normal
temperatures in June 1816. The typical synoptic situation giving southwest winds over
Newfoundland in June is a ridge of high pressure pushing northward from the Bermuda High.

In northern Labrador and Hudson Strait, Moravian records (Newell 1983) and Hudson's Bay
Company records (Catchpole and Faurer 1985; Teillet 1988) indicate severe ice conditions with
considerably delayed clearing dates. Newell (1983) states that in 1816 it was " likely that the sea
ice had not completely cleared the coast by the start of the next [ice] season". Analysis of sea-ice
clearing in this region based on satellite imagery for 1964-74 (Crane 1978) demonstrates that late
clearing dates are associated with an increased frequency of northerly winds. Catchpole and
Faurer (1985), investigating sea-ice conditions in Hudson Strait during 1816 using logs from
Hudson's Bay Company ships, also found evidence for an increased frequency of northerly winds
during the summer.

Besides providing valuable information regarding the offshore ice conditions, the Moravian
mission reports also provide some indication of the weather experienced at the stations. The
following remarks regarding the summer of 1816 at Okkak follow a description of the severity
of the winter: "In spring, the frost continued so severe, that we could not work in our gardens
at the proper time, and consequently expect but a poor crop of vegetables this year, for the whole
summer season has been cold and dry" (Periodical Accounts, Vol. VI, p. 265). The following
reports from the Moravian missions in southwestern Greenland suggest different conditions on
the other side of the Labrador Sea: "It rains almost incessantly, and if it even ceases for a day,
yet the heavens are overcast... I must say that for these four months past, we have not had one
day on which the sun has shone throughout the whole day" (Periodical Accounts, Vol. VI,
p. 452). An analysis of conventional meteorological data collected at the Labrador Moravian
stations in the 1880s and 1890s suggests that in June cold dry conditions are associated with north
or northwest winds and lower air pressures; both of which would occur with a mean low-pressure
centre to the east and lows tracking well south of the area. The wet conditions in southwestern
Greenland indicate that this area was near or just east of the main low-pressure centre.

In summary, the data presented indicate that during the spring and summer of 1816 a mean centre
of low pressure was situated in the Labrador Sea with a trough extending north into Davis Strait
(Figure 3). At the same time the main track of low-pressure systems was across southern
Labrador and into the Labrador Sea. This pattern would give the north to northwest winds and
cold/dry conditions in Labrador and the wet conditions in southwestern Greenland. South of the
storm track, southeastern Newfoundland was under the influence of the Bermuda High. The
temperature pattern for June 1816 has very cold conditions in northern Labrador and normal to
above normal temperatures in Newfoundland.

248

Wind Direction Deg-

Figure 2: Air temperature versus wind direction for St. John's, Newfoundland. Based on data for June
1971-87, supplied by the Atmospheric Environment Service, Scientific Service Unit,
St. John's.

249

Modern Analogues

To provide a check on the proposed circulation pattern for 1816 and to give more detail on the
nature of the circulation, modern analogues for the temperature pattern observed in June 1816
were selected, and their circulation patterns compared to that proposed. The criteria used were
below-normal temperatures in northern Labrador and normal or above-normal temperatures in
southeastern Newfoundland in June. Monthly temperature patterns were obtained from maps in
Environment Canada publications (Climatic Perspectives and Monthly Record). During the 30-
year period 1958-87, five years had June temperatures that met the criteria (1969, 1971, 1972,
1978 and 1986; Figure 4).

All of the years selected as analogues had below-normal June temperatures at Churchill,
Manitoba, on the west coast of Hudson Bay. This pattern agrees well with conditions in 1816
when temperatures at Churchill were considerably below normal (Catchpole 1985). All of these
years except 1986 had cool to very cold conditions in central England; in fact, June 1971 and
1972 were colder than June 1816 (Manley 1974). The opposition of temperatures in
Newfoundland and England agrees with the Burroughs' (1979) finding of an inverse relationship
between temperatures in the two areas. The agreement between conditions in the five years
mentioned above and 1816 is not as strong when conditions in New England are considered. Only
two of the five years (1972 and 1986) had below-normal June temperatures at Boston: however,
in all five years below-normal temperatures reached some part of New England.

In all but one of the five years considered, the mean centre of low pressure in the North Atlantic
was near its normal position, over or near the Labrador Sea. The exceptional year was 1972,
when the low was south of Iceland. However, in all cases mentioned, the circulation was more
intense than normal. Of the five years considered, the circulation patterns in 1969 and 1971 seem
most unlike that for 1816. In these cases the surface winds in northern Labrador had a strong
southerly component - totally unlike 1816. Perhaps the surface temperatures indicated for
northern Labrador during these years (based on data from surrounding stations) are in error, and
the true temperatures were higher. In the case of 1972, while the temperature pattern matches that
for June 1816 in England, New England, Hudson Bay, Newfoundland and Labrador, the nature
of the circulation does not fit the proposed pattern for the eastern Atlantic/ western Europe sector
as proposed by Kelly et a\. (1984). In fact the circulation for June 1972 is totally different from
the conditions that usually produce below-normal June temperatures over England (Perry 1972).

The two remaining years (1978 and 1986) both have deeper-than-normal low-pressure centres
over the Labrador Sea; however, in 1978 a high-pressure centre occurring over Hudson Bay was
absent in 1986. Wilson (1985) proposed that such a high was centred over Hudson Bay during
the summer of 1816. The 700-mb circulations in both of these years have some similarities and
some major differences. Both years have above normal 700-mb heights southeast of
Newfoundland. This feature also occurs during the three other years selected, and is likely related
to above-normal temperatures in Newfoundland. Farther north the 700-mb patterns for the two
years are different. In June 1986 there is a trough along the Labrador coast with the largest
negative height departures over the mid-Labrador coast, whereas in June 1978 the trough is
farther west, the greatest negative height departures being over Foxe Basin.

250

Figure 3: Mean June sea-level pressure: (A) estimated for 1816 and (B) present-day normals.

251

Figure 4: Mean sea-level pressure (mb) and regions with below-normal air temperature for June: 1969,
1971, 1972, 1978 and 1986.

252

Analysis of Environment Canada (Atmospheric Environment Service) ice charts indicate that
clearing dates for the Labrador coast were later than normal in both 1978 and 1986; however,
neither year represented record conditions. Ice conditions at the end of June 1978 were more
severe than at the end of June 1986, but the rate of retreat during the month of June was greater
in 1978 than in 1986. Since this analysis only considered conditions in June, it is not surprising
that ice conditions were not as severe as in 1816. Conditions earlier in the spring, and the strong
northerly flow in July 1816 indicated by Lamb and Johnson (1966), likely played an important
role in the exceptional 1816 ice conditions.

Summary

Comparison of circulation and temperature patterns for June 1978 and 1986 with the proposed
pattern for June 1816 (Figure 3) indicates that they are in general agreement. For example, all
three maps have a deep low-pressure centre in the Labrador Sea. However, apparently the
northerly circulation in 1816 must have been more vigorous than in 1978 or 1986 to give the
lower temperatures reported. This would require that the low-pressure centre in the Labrador Sea
be deeper than in either of those years. The actual pattern for June 1816 likely combined features
of both June 1978 and 1986. This pattern also agrees with the North Atlantic circulation for July
1816 proposed by Lamb and Johnson (1966).

The occurrence of a circulation pattern such as the one proposed for June 1816 without outside
forcing (such as volcanic cooling) does not seem unrealistic in light of the variability
demonstrated in the five analogues considered in this study. Perhaps such an occurrence is
especially likely considering that in 1816 the northern hemisphere was experiencing the last stages
of the Little Ice Age, a period when such circulation patterns would have been more common.
The main difficulty with this argument is that data from other sources demonstrate that conditions
during July and August 1816 were equally unusual. A long-term data set of sea-ice conditions
for the Labrador Sea that I am currently developing may assist in determining how the summer
of 1816 compares with modern conditions and with other summers in the nineteenth century.

References

Briffa, K.R., P.D. Jones and F.H. Schweingruber. 1988. Summer temperature patterns over
Europe: a reconstruction from 1750 A.D. based on maximum latewood density indices of
conifers. Quaternary Research 30:36-52.

Burroughs, W.J. 1979. An analysis of winter temperatures in central England and Newfoundland.
Weather 34:19-23.

Catchpole, A.J.W. 1985. Evidence from Hudson Bay region of severe cold in the summer of
1816. In: Critical Periods in the Quaternary Climatic History of Northern North America.
Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus 55:121-146.

Catchpole, A.J.W. and M.-A. Faurer. 1985. Ships' logbooks, sea ice and the cold summer of
1816 in Hudson Bay and its approaches. Arctic 38:121-128.

Crane, R.G. 1978. Seasonal variations of sea ice extent in the Davis Strait-Labrador Sea area and
relationships with synoptic-scale atmospheric circulation. Arctic 31:434-447.

253

Forsey, R. and W.H. Lear. 1987. Historical catches and catch rates of Atlantic Cod at
Newfoundland during 1677-1833. Department of Fisheries and Oceans, Canadian Data
Report of Fisheries and Aquatic Sciences No. 662:1-52.

Kelly, P.M., T.M.L. Wigley and P.D. Jones. 1984. European pressure maps for 1815-16, the
time of the eruption of Tambora. Climate Monitor 13:76-91.

Lamb, H.H. 1977. Climate: Present, Past and Future. Vol. 2; Climatic History and the Future.
Methuen, London. 835 pp.

Lamb, H.H. and A.I. Johnson. 1959. Climatic variation and observed changes in the general
circulation, Parts I and II. Geografiska Annaler 41:94-134.

. 1961. Climatic variation and observed changes in the general circulation, Part III.

Geografiska Annaler 43:363-400.

. 1966. Secular variations of the atmospheric circulation since 1750. Great Britain,

Meteorological Office, Geophysical Memoirs 110:1-57.

Ludlum, D.M. 1966. Early American Winters 1604-1820. American Meteorological Society,
Boston.

Manley, G. 1974. Central England temperatures: 1659-1973. Quarterly Journal of the Royal
Meteorological Society 100:389-405.

Newell, J. P. 1983. Preliminary analysis of sea-ice conditions in the Labrador Sea during the
nineteenth century. In: Climatic Change in Canada 3. C.R. Harington (ed.). Syllogeus
49:108-129.

Perry, A.H. 1972. June 1972 - the coldest June of the century. Weather 27:418-422.

Scoresby, W., Jr. 1820. An Account of the Arctic Regions with a History and Description of the
Northern Whale-Fishery. Reprinted in 1969 by Augustus M. Kelley, Publishers, New
York.

Stommel, H. and E. Stommel. 1979. The year without a summer. Scientific American
240:176-186.

Teillet, J.V. 1988. A reconstruction of summer sea ice conditions in the Labrador Sea using
Hudson's Bay Company ships' log-books, 1751-1870. Unpublished M.A. thesis,
University of Manitoba, Winnipeg. 161 pp.

Wilson, C.V. 1983. The summer season along the east coast of Hudson Bay during the nineteenth
century. Part II: The Little Ice Age on eastern Hudson Bay: summers at Great Whale, Fort
George, Eastmain, 1814-1821. Canadian Climate Centre, Downsview, Report No. 83-9.

. 1985. The Little Ice Age on eastern Hudson/James Bay: the summer weather and climate

at Great Whale, Fort George and Eastmain, 1814 to 1821, as derived from Hudson's Bay
Company records. In: Critical Periods in the Quaternary Climatic History of Northern
North America. Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus
55:147-190.

254

Spatial Patterns of Tree-Growth Anomalies from the North American
Boreal Treeline in the Early 1800s, Including the Year 1816

Gordon C. Jacoby, Jr.1 and Rosanne D'Arrigo1
Abstract

Tree-growth anomalies based on 24 temperature-sensitive white spruce chronologies from boreal
treeline sites in North America are mapped and analyzed for the interval 1805-24. This interval
includes the volcanic eruption of Tambora in 1815 and the unusual "year without a summer" in
1816. The first few decades of the 1800s also are concurrent with a series of low-amplitude
cycles in sunspot number which have been suggested as contributing to unusually cooler
conditions during this time. It is inferred from the tree-ring data that climatic changes following
the Tambora eruption influenced the North American boreal forest in different areas at different
times from 1816 to 1818, with the coldest regional temperatures appearing to have occurred in
the year 1816 in easternmost Canada. The series of anomaly maps provided here help to clarify
the spatial patterns of climatic changes in remote northern regions during this extreme cooling
event.

Introduction

The climatically unusual "year of no summer" (1816) was primarily documented as such by
observers in Europe and eastern North America (Landsberg and Albert 1974; Stommel and
Stommel 1983; Stothers 1984; Briffa et al. 1988). Cold air masses invaded most areas of Europe
and the settled, eastern regions of North America (Stommel and Stommel 1983; Stothers 1984;
Briffa et al. 1988). There is little documentation on weather variations for western North America
about the time of the Tambora eruption (1815), and much of the documentation on weather
variations for other parts of the world is only recently being brought into full consideration (e.g.,
Legrand and Delmas 1987, concerning evidence of the Tambora eruption in Antarctic ice-core
data; also other papers in this volume). Overall, little is known regarding the regional -scale
climatic variations following many volcanic eruptions (including Tambora), and it is likely that
hemispheric-scale studies demonstrating a general cooling effect may obscure warming in some
areas (Lough and Fritts 1987) that may result from changes in large-scale atmospheric dynamics
following major eruptions (Hansen et al. 1978; Schneider 1983).

It has been hypothesized that volcanism can strongly influence climate, causing cooler
temperatures (e.g., Lamb 1970; Mass and Schneider 1977; Sear etal. 1987; Bradley 1988). The
mechanism is not thoroughly understood but the common theory is that stratospheric sulphate
particles partially reflect and absorb incoming radiation, heating the stratosphere. This heat does
not reach the troposphere, which then becomes cooler (e.g., Hansen et al. 1978). Although
empirical modeling studies (Hansen et al. 1978) and superposed epoch analyses (Mass and
Schneider 1977; Sear et al. 1987; Skinner, this volume) indicate a cooling of a few tenths of an

Tree-Ring Laboratory, Lamont-Doherty Geological Observatory of Columbia University, Palisades, New York
10964, U.S.A.

255

degree following major eruptions, this cooling is within the level of natural climatic variability.
Hence a link cannot be unequivocally proven, but there is strong evidence for a cause and effect
relationship (Sear et al. 1987).

The eruption of Tambora in April 1815 was one of the greatest volcanic events of recent centuries
(Simkin et al. 1981) . Because studies indicate (Sear et al. 1987) that southern hemisphere
eruptions would cool northern hemisphere temperatures after a lag of about six months to a year,
a response to this event at northern mid- to high-latitudes would not be expected to occur until
late 1815-16. An unusually cold summer in 1816 is attributed by many to the effects of this event
but the overall climatic anomalies of the period are more complex than a single event-response
phenomenon. One phenomenon to be considered in the complexity is that the early 1800s were
notable for a reduction in solar sunspot amplitudes, possibly reflecting solar irradiance changes
(Lean and Foukal 1988; Kuhn et al. 1988) that may have contributed to a cooling during the
early decades of the 1800s (Eddy 1977, this volume).

Perspective of This Study

The method used here to examine the climate of 1816 is the study of old-aged trees growing
along the northern boreal forest zone of North America. All of the data are from white spruce
[Picea glauca (Moench) Voss] near the forest-tundra transition zone. These trees primarily
respond to summer temperature, with a secondary response to fall and spring temperature-related
conditions (Jacoby and D'Arrigo, in press; Scott et al. 1988; Jacoby and Ulan 1982; Cropper
1982; Jacoby and Cook 1981; Garfinkel and Brubaker 1980). The average position of the Polar
Front in summer largely coincides with the location of the northern treeline (Bryson 1966). Tree
growth in this region can therefore be expected to record frontal shifts, extensive outbreaks of
polar air masses and circulation changes influencing thermal conditions (e.g., Scott et al. 1988).
Thus tree-ring data can provide useful information on climatic response following the 1815
eruption (and the early 1800s in general) in the northern boreal region, including the western
region of North America where other data sources are scarce.

We have examined 24 time-series of absolutely dated tree-ring width indices (chronologies)
throughout the region. Each time-series is usually based on about 10 trees with multiple cores
(radii) from each tree. Some of these chronologies contain low-frequency response to climate,
whereas others only preserve higher-frequency response. For compatibility and intercomparison,
all chronologies were prewhitened (to remove low-frequency variation) and normalized. This
procedure is appropriate since, in this case, we are evaluating variations in year-to-year response
over a period of a few decades. Maps (Figures 1, 2) display the departures from the mean for
each chronology location during the early 1800s (1805-24).

Distribution of Anomalies

For 1816, Figure 1 shows substantial negative departures (reflecting reduced radial growth/colder
temperatures) in eastern Canada. These are the greatest anomalies for the period under review
(1805-24). However, the rest of Canada and Alaska shows no such severe cooling. Northern
Alaska was fairly cold (as in some other years), but central and western Canada are close to
average or above for the year. The colder eastern region agrees well with reports and records
from eastern Canada and the United States. For example in the northern region, Catchpole and
Faurer (1983) demonstrate that the duration of westward passage of Hudson's Bay Company ships
was the longest (54 days compared to a mean of 17.7 days) of the entire 1751-1870 record,

256

representing severe sea-ice conditions. The authors (see also Catchpole, this volume) suggest that
these conditions could be explained by enhanced meridional flow of arctic air masses over eastern
North America at this time. Records from the eastern United States, do not show a continuously
cold summer (Stommel and Stommel 1983; Baron, this volume). There were three distinct
outbreaks of extremely cold air from the North at different times during the summer. These
outbreaks had serious negative effects on food crops during the growing season (Stommel and
Stommel 1983). Such records appear to support the theory of increased Arctic air flow over this
region in 1816. Schneider (1983) suggests that since the cooling (about 3°C) would not have been
sufficient to explain the documented frosts that occurred, a dip in the Jetstream and blocking of
the mid-latitude westerlies could have contributed to the adverse conditions. He suggests that
conditions to the west (about one-half wavelength away) of eastern North America would have
been unusually warm if this had been the case. Our results support this contention since
conditions in central Canada, although not unusually warm, do not demonstrate the pronounced
cooling found in the east (Figure 1).

Figure 1: Tree-growth anomaly map for 1816. The growth departures are based on prewhitened and
normalized tree-ring width indices for 24 white spruce chronologies from near the boreal
treeline of North America.

To place this year in context, we review the years preceding and after 1816 beginning with 1805
(Figure 2). The three years of 1805-07 show little in the way of extreme cold temperatures.
Except for southeastern Alaska, most other regions are near or above normal, and eastern Canada
is substantially above normal. Then in 1808 colder temperatures prevailed in the Hudson Bay
region, and Alaska was warm. In 1809 the Hudson Bay region was less cold but Alaska became
cold. The distribution of regions of warmer or cooler temperatures correspond roughly to the
configuration of the longwave pattern in the atmosphere. There is approximately one wavelength
across the North American quadrant for a four-wave pattern, western Alaska to eastern Canada
being slightly over 90° of longitude (Chang 1972).

257

258

ure 2 (cont'd):

Figure 2 (cont'd):

260

Figure 2 (cont'd):

During 1810 through 1812 the main features of the maps are a cooling over eastern Alaska, the
Yukon Territory and the western Northwest Territories, and in Labrador an alternating
warm-cool -warm sequence. An indication of very warm conditions in 1813 over the Northwest
Territories is followed by a reversal to quite cold conditions for Alaska, especially southeastern
Alaska, and all of western Canada in 1814 accompanied by a warming in Labrador. The cooler
conditions continue for the western region in 1815. Northern Alaska is fairly cold in 1816, but
the most severe cold is restricted to eastern Canada (see also Figure 1). As noted above, central
and western Canada are not unusually cold during 1816. More severe cold does not reach western
Canada and eastern Alaska until 1817 when the anomalies are more negative than other years of
the period, although 1809 is quite cold. The eastern region is still cold but recovering toward
normal. In 1818, the coldest conditions occur in the Northwest Territories. Again these are the
greatest negative anomalies for this area, although 1814 and 1821-22 are also cold. By 1819,
almost all extreme negative anomalies are gone from the map region.

Extreme western Alaska and eastern Canada are warm in 1820 but cold pervades much of the
entire map region during 1821 and 1822, except for Labrador in 1822. Alaska warms in 1823
but there is a return to cold temperatures in 1824 in both western Canada/Alaska and in eastern
Canada.

In summary, there were significantly cold conditions in some northern areas before 1816. After
the volcanic event, extreme cold affected all of the map region at different times until 1818. Cold
temperatures pervaded some areas after 1818, and 1824 was fairly cool throughout most of the
region.

Discussion and Conclusions

Our results show the spatial patterns of tree-growth anomalies from the North American boreal
treeline during the anomalous period of the early 1800s, with an emphasis on the year 1816. The
data provide added spatial coverage of western Canada and Alaska in relation to the climatic
response following the Tambora (1815) event. In agreement with other studies, apparently the
unusual cold in eastern North America may in part have resulted from meridional flow of cold
Arctic air across this region. By contrast, conditions in western and central Canada (about a half
wavelength to the west) were moderate in 1816 (Schneider 1983). This reflects a possible shift
in atmospheric circulation which may or may not be directly linked to the volcanic event.

Records of climatic conditions in the early 1800s in other regions of the globe are rather
fragmentary, as discussed, and the cooling in the early 1800s was probably not synchronous
globally (see the Workshop section, this volume). Figure 3 shows a reconstruction of northern
hemisphere annual temperatures (from Jacoby and D'Arrigo 1988) based on North American
boreal tree-ring data, indicating a cooling during this interval that persisted for several decades.
On a more regional scale, the cooling in Europe is well documented (Stommel and Stommel
1983; Briffa et al. 1988). Specifically, a sharp lowering of temperature is seen in England and
central Europe from 1812-20 in reconstructed summer temperatures based on tree-ring density
data (Briffa et al. 1988). Records from China and Japan (Stommel and Stommel 1983) do not
indicate unusually cold conditions1. A detailed compilation of spatial temperature data during this
time interval is clearly needed.

But see Zhang et al. and Huang, this volume (editor).

262

RECONSTRUCTED NORTHERN HEMISPHERE TEMPERATURE DEPARTURES

1670

1710

1750

1790

1830

1870

1910

1950

1990

YEARS

Figure 3: Reconstruction of annual northern hemisphere temperatures from 1671 to 1973 based on high
latitude tree-ring data from North America. Temperature departures from 1974-87 from Hansen
and Lebedeff, 1987, 1988 [see Jacoby and D'Arrigo (in press)]. Note the abrupt cooling in the
early 1800s.

The detection of a direct cause and effect signal due to volcanism is difficult, in part due to the
influence of other forcing functions on the climatic system. These include unusual (diminished)
solar fluctuations that also occurred in the early 1800s (Eddy 1977). El Nino events occur on the
same time scale as volcanic eruptions (largely high frequency) and can obscure their signal, as
can random climatic variations (Robock 1981). Modeling studies (e.g., Gilliland and Schneider
1984; Robock 1981) show good agreement between model estimates (based on volcanic indices)
and actual temperature data but other forcings must also be considered. Finally there are many
complicating factors for individual eruptions [e.g., season and latitude of eruption, height and
chemistry of ejecta, state of atmospheric circulation at time of eruption (Lamb 1970; Lough and
Fritts 1987)] which complicate attempts to detect a common event-response pattern.
Improvements in understanding volcanic forcings are necessary for isolating effects of other
forcings such as C02.

The oft-applied term "year of no summer" for 1816 is obviously a misnomer in the context of
the northern boreal forests of North America. To understand climatic change in Canada and the
rest of North America, it is necessary to move away from this oversimplification and study spatial
and temporal differences and dynamics of the early 1800s, as these decades are a time of

263

substantial climatic variation. Here we have provided a series of maps from the early 1800s that
help clarify the spatial patterns of climate at remote high-northern latitudes during this interval,
and which may be useful in determining causes of and responses to such extreme climatic events.

Acknowledgements

This research was supported by the Climate Dynamics Division of the National Science
Foundation, under grants ATM85-15290 and ATM87-16630. We thank J. Hayes and
W. Ruddiman for helpful reviews, and the Canadian Forestry and Atmospheric Environment
services for technical assistance. Lamont-Doherty Geological Observatory Contribution No. 4566.

References

Bradley, R.S. 1988. The explosive volcanic eruption signal in northern hemisphere continental
temperature records. Climatic Change 12:221-243.

Briffa, K., P.D. Jones and F.H. Schweingruber. 1988. Summer temperature patterns over
Europe: a reconstruction from 1750 A.D. based on maximum latewood density indices of
conifers. Quaternary Research 30:36-52.

Bryson, R.A. 1966. Air masses, streamlines, and the boreal forest. Geographical Bulletin
8:228-269.

Catchpole, A.J.W. and M.A. Faurer. 1983. Summer sea ice severity in Hudson Strait,
1751-1870. Climatic Change 5:115-139.

Chang, J.H. 1972. Atmospheric Circulation Systems and Climates. Oriental Publishing Co.,
Honolulu. 326 pp.

Cropper, J. P. 1982. Climate reconstructions (1801 to 1938) inferred from tree-ring width
chronologies from the North American Arctic. Arctic and Alpine Research 14:223-241.

Eddy, J. A. 1977. Climate and the changing sun. Climatic Change 1:173-190.

Garfinkel, H.L. and L.B. Brubaker. 1980. Modern climate-tree growth relationships and climatic
reconstruction in subarctic Alaska. Nature 286:872-874.

Gilliland, R.L. and S.H. Schneider. 1984. Volcanic, CO2 and solar forcing of northern and
southern hemisphere surface temperatures. Nature 310:38-41.

Hansen, J.E., W.C. Wang and A. A. Lacis. 1978. Mount Agung eruption provides test of a
global climatic perturbation. Science 199:1065-1068.

Jacoby, G.C. Jr. and E.R. Cook. 1981. Past temperature variations inferred from a 400-year
tree-ring chronology from Yukon Territory, Canada. Arctic and Alpine Research
13:409-418.

Jacoby, G.C. Jr. and L.D. Ulan. 1982. Reconstruction of past ice conditions in a Hudson Bay
estuary using tree-rings. Nature 298:637-639.

264

Jacoby, G.C. Jr. and R. D'Arrigo. 1988. Reconstructed northern hemisphere annual temperature
since 1671 based on high latitude tree-ring data from North America. Climatic Change (in
press).

Kuhn, J.R., K.G. Libbrecht and R.H. Dicke. 1988. The surface temperature of the sun and
changes in the solar constant. Science 242:908-911.

Lamb, H.H. 1970. Volcanic dust in the atmosphere; with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society of London
255:425-533.

Landsberg, H.E. and J.M. Albert. 1974. The summer of 1816 and volcanism.
Weatherwise 27:63-66.

Lean, J. and P. Foukal. 1988. A model of solar luminosity modulation by magnetic activity
between 1954 and 1984. Science 240:906-908.

Legrand, M. and R.J. Delmas. 1987. A 220-year continuous record of volcanic H2S04 in the
Antarctic Ice Sheet. Nature 327:671-676.

Lough, J.M. and H.C. Fritts. 1987. An assessment of the possible effects of volcanic eruptions
on North American climate using tree-ring data, 1602 to 1900 A.D. Climatic Change
10:219-239.

Mass, C. and S. Schneider. 1977. Statistical evidence on the influence of sunspots and volcanic
dust on long-term temperature records. Journal of Atmospheric Science 34:1995-2004.

Robock, A. 1981. A latitudinally dependent volcanic dust veil index and its effect on climate
simulations. Journal of Volcanology Geothermal Research 11:67-80.

Schneider, S. 1983. Volcanic dust veils and climate: how clear is the connection? - an editorial.
Climatic Change 5:111-113.

Scott, P. A., D.C.F. Fayle, C.V. Bentley and R.I.C. Hansell. 1988. Large-scale changes in
atmospheric circulation interpreted from patterns of tree growth at Churchill, Manitoba,
Canada. Arctic and Alpine Research 20:199-211.

Sear, C.B., P.M. Kelly, P.D. Jones and CM. Goodess. 1987. Global surface-temperature
responses to major volcanic eruptions. Nature 330:365-367.

Simkin, T., L. Seibert, L. McClelland, W.G. Melson, D. Bridge, C.G. Newhall and J. Latter.
1981. Volcanoes of the World. Smithsonian Institution, Washington, D.C.

Stommel, H. and E. Stommel. 1983. Volcano Weather: The Story of 1816: the Year Without a
Summer. Seven Seas Press, Newport. 177 pp.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science
224:1191-1198.

265

Early Nineteenth-Century Tree-Ring Series from Treeline Sites
in the Middle Canadian Rockies

B.H. Luckman1 and M.E. Colenutt1

Abstract

Preliminary data from tree-ring series at treeline sites in the Canadian Rockies are evaluated for
evidence of anomalies associated with the 1815 eruption of Tambora. Three maximum density
and nine ring-width chronologies (six Picea engelmannii, one each for Abies lasiocarpa, Larix
lyallii and Pinus albicaulis) are presented, covering 1780-1860. The absence of narrow or light
latewood marker-rings associated with 1816 or 1817 indicates that there is no distinctive tree-ring
signal associated with the Tambora event at these sites. Most records do however contain a sharp
decrease in ring widths during the 1810-20 decade, similar to that reported from latitudinal
treeline sites elsewhere in North America, and which appears to be associated with an abrupt
deterioration of climate initiated several years prior to the Tambora eruption.

Introduction

Evaluation of the spatial extent of climatic anomalies associated with the eruption of Tambora in
1815 requires assessment of proxy-data series throughout North America. The annual resolution
of tree-ring series, combined with strong relationships between ring characteristics and climate,
potentially provide a powerful tool to accomplish this goal. Here we present data from
preliminary tree-ring chronologies for 1780-1860 at several treeline sites in and adjacent to Banff
and Jasper national parks in the Canadian Rocky Mountains (Figure 1). These data are examined
to see whether significant anomalies are present in the 1815-17 period that could be attributed to
climatic effects associated with the Tambora eruption.

The principal direct climatic effect associated with volcanic eruptions is a reduction in solar
radiation received at the surface because of stratospheric dust veils (Lamb 1970). This often
results in cooler summers, and the spatial extent and severity of this effect depends on the
magnitude, timing and nature of the eruption. LaMarche and Hirschboeck (1984) have
demonstrated a strong relationship between the presence of frost rings in the Bristlecone pine
chronology from treeline sites in the White Mountains of California and major volcanic eruptions.
These data include a frost-ring date of 1626 B.C. for the eruption of Santorini in Greece (which
destroyed the Minoan civilization of Crete, and probably had global climatic effects). Baillie and
Munroe (1988) show that Irish bog oaks had very narrow rings in the 1620s (B.C.), which appear
to confirm this result. In Canada, Filion et al. (1986) have shown that light latewood rings from
black spruce chronologies in northern Quebec correspond with periods 0-2 years after major
volcanic eruptions. In these records the 1816-17 rings have light latewood in 75% of the series
studied, and 1784 (the year following the Laki eruption) is also a prominent marker ring. Parker
(1985) and Jacoby et al. (1988) also note the exceptional nature of the 1816 and 1817 rings in
white-spruce chronologies on the eastern shores of Hudson Bay. The severe climate of these two
summers is amply demonstrated by several papers in this volume.

Department of Geography, University of Western Ontario, London, Ontario N6A 5C2, Canada.

266

120°W

r ALBERTA

Figure 1: Location of the main study sites.

Parker (1985) evaluated selected tree-ring series from western and central Canada to determine
whether he could detect a signal associated with the eruptions of Tambora or Krakatau (1888).
He used ring-width and densitometric data from 135 trees (four different species) at 15 sites
between Vancouver Island and Hudson Bay. The data were aggregated into six regional
chronologies, and indexed data were used to compare the eruption year with groups of three years
before and after the eruption. Only one site, Cri Lake on Hudson Bay, showed a significant
growth reduction following the Tambora eruption.

These results suggest that, under certain conditions, a volcanic signal can be detected via its
influence on climate, and thereby on tree-ring characteristics. Many authors have demonstrated
strong relationships between tree-ring width or density series and summer temperatures -
particularly at treeline sites (e.g., Parker and Henoch 1971; Luckman et al. 1985; Jacoby and
Cook 1981; Jacoby et al. 1988). It would be anticipated, therefore, that trees at these sites would
be particularly sensitive to reductions in summer insolation, and therefore most likely to record
evidence of dust-veil-related volcanic effects. As most of the montane sites used by Parker (1985)
are well below treeline, we decided to evaluate treeline records from the Rockies to see whether
the 1815-17 record contained any distinctive signal that could be attributed to the effects of the
Tambora eruption.

267

Sample Sites

Nine preliminary living-tree ring-width chronologies are available from our tree-ring studies in
the Canadian Rockies (Table 1, Figure 1): eight are from treeline sites and five (Robson,
Bennington and Icefields/Athabasca sites) are adjacent to Little Ice Age terminal moraines. Six
of these chronologies utilize Engelmann spruce (Picea engelmannii) because it is the most
ubiquitous, long-lived tree at treeline in this area. Single chronologies for alpine larch (Larix
lyallii), alpine fir {Abies lasiocarpa) and whitebark pine (Pinus albicaulis) are also used.
Tree-ring densitometric data are also available for three of these sites.

The Robson site (Figure 2) is an isolated stand of spruce on a low bedrock knoll overlooking an
inactive outwash fan from Robson Glacier. At its Little Ice Age maximum position, Robson
Glacier advanced against the upvalley side of the knoll and built a terminal moraine along its
crest. Heusser (1956) estimated the date of formation of the three outermost moraines of Robson
Glacier as 1787, 1801 and 1861 based on tree-ring sampling and allowing a 12-year ecesis
interval. The oldest tree in the stand outside the moraine is just over 400 years old, i.e., it
predates the maximum glacier advance by about 200 years. The trees were sampled in 1981 and
1983. Preliminary results are given by Watson (1983): the results presented here use both data
sets.

Figure 2: The Robson Glacier site, view east
from Adolphus Lake, Alberta
(foreground) toward Rearguard
Mountain and Mount Robson
(snow-covered, top right). The
sampled stand (a) is visible with the
lighter-toned Little Ice Age moraine
complex of the Robson Glacier
extending from left to right across
the middle ground behind the trees.

268

"o
c

'—

r-i

■

l/">

r-4

t*«

S3 p

Q
o£ a;

Q
X

£ £ £

o£ c£ Q£

Sis

w w <

w J J

Ph < J

a
-J

>
5

O m

On vo

no r-~

-J

—J

o
o

.3 z

4>
00

o o
— cs

ON 00

.2 c
c

C/J

o
05

-C

s
a

2
c

W
II

a
x

<u c

<->

s:
s:
a

< £

uj
a.

269

The Bennington site is an open grown, almost pure stand of whitebark pine growing on a coarse
talus and bedrock slope overlooking the lateral moraine of Bennington Glacier (Figure 3) which
is dated to about 1700 and 1825 by dendrochronology (MacCarthy 1985). This stand contains a
number of very old trees, the oldest of which has a pith date of 1 1 12 A.D. at breast height, and
is thought to be the oldest whitebark pine in Canada (Luckman et al. 1984). Chronology
development at this site is incomplete, and data given are for the four trees for which ring width
and densitometric data are presently available for the 1780-1860 interval. The spruce stand at this
site is on the lower slope, slightly downvalley of the area shown in Figure 3. The tree-ring series
from this site show high tree-to-tree variability due to rockfall disturbance of the site (Watson
1983). The results are included here solely for comparison with spruce chronologies at other
sites.

The Athadome (Figure 4) and Icefield (Figure 5) sites are both adjacent to the Athabasca Glacier
on opposite sides of the Sunwapta Valley, less than a kilometre apart. The Athadome site has a
unique microclimate because it lies between the lateral moraines of Athabasca and Dome glaciers
and was almost completely surrounded by and below the level of the adjacent ice surface about
1714 (Heusser 1956) and between approximately 1840 and 1920 (Luckman 1988). By contrast,
the Icefield site is a well drained lower valley side slope just beyond the outer limits of the
Athabasca Glacier. Both chronologies are Engelmann spruce, but sampling in 1980 and 1981
(Luckman 1982) indicated that some trees at the Icefield site were considerably older. Intensive
sampling at this site in 1982 provided the present chronology (Jozsa et al. 1983), which is based
on trees with a mean age of over 500 years (Table 2) and includes the oldest known Engelmann
spruce (Luckman et al. 1984).

The Lake Louise site is the only non-treeline site presented here. It occurs in the lower subalpine
forest about 300 m below treeline on a valley side bench overlooking Lake Louise townsite
(Hamilton 1984). This site was the closest Engelmann spruce stand to the meteorological station
at Lake Louise, and was used to explore climatic tree-ring relationships for this species (Luckman
et al. 1985). The Larch Valley site (Figure 6) is at treeline, some 10 km south of Lake Louise.
It is about 2 km from the Wenkchemna Glacier, and considerably above it on a broad valley side
bench overlooking the main valley. Chronologies were developed for three species in the same
stand at this site because of difficulties in crossdating the larch record which has several periods
with very tight or missing rings (Colenutt 1988). This larch chronology is the best-replicated and
most sensitive (mean sensitivity 0.38) of those discussed here. The Larch Valley and Lake Louise
chronologies are also less likely to show local climatic effects from adjacent glaciers than the
other chronologies reported here.

Chronologies for most sites were developed by standard methods using the Laboratory of
Tree-Ring Research (Tucson) programs INDEX and SUMAC (Graybill 1982). However,
chronologies for the Icefield and Lake Louise sites were developed by Forintek, Vancouver using
a 99-year running mean to remove the growth trend (Parker et al. 1981). Therefore some of the
longer-frequency trends in these chronologies have been removed resulting in a lower amplitude
of response (Luckman et al. 1985).

Results

The results from the nine indexed ring-width chronologies are shown in Figure 7, and high-pass
filter data (Fritts 1976) from these series are presented in Figure 8. The indexed values for
1810-20 are listed in Table 2 with some summary statistics for the chronologies used. These

270

CS
00

OS
00

00
00

>/->

cs
oo

o —
— o

o
oo

Oft

r-
as

CS
O

SO
CS

00
CS

CS
OS

SO
Os

OS
OS

"~! so

C/5

so
00

vO
OS

so
co

cs
CO

m
cs

00
CS

W
Ou

OS
00

00
OS

Os
CI

I 2

< £

OS
Os

00
00

00
Os

CL,

r—

»/■>
r-

c->
so

SO
SO

r»

OS
IT)

r»

OS
SO

o
r~

cs
r-

t-~
r>

CS
OS

o
r-

W>
OS

c->

ley

>>
j)

Va]

J3 3 3

a
a

s:
s:
a

•S

LU
CU

•Si

•x

-3

C
3

E
u

C T3

c C

CS t)
v.

o "V

" cs

O 00

E 5

Z v2

271

Figure 3: The Bennington Pine site, view north across the north lateral of Bennington Glacier (July 1986);

the main sampled area in the centre of the photo contains many standing snags and trees over
500 years old. The spruce and fir flanking the moraine are much younger in age. The ridgecrest
is about 1000 m above the valley floor.

Figure 4: View south across the forefield of the Athabasca Glacier to the lateral moraines of Athabasca
Glacier (left) and Dome Glacier (right). The Athadome site is near the small group of trees at
(a), the Icefields site is visible in the foreground, north of the Icefields Parkway.

272

Figure 5: View west across the forefield of the Athabasca Glacier to the Icefields site (area b). Note the
well-marked trimline denoting the Little Ice Age limit (about 1842 A.D.) of the Athabasca
Glacier against the slope.

Figure 6: Part of the main stand sampled at Larch Valley. View north toward Sentinel Pass with Mount
Temple (right).

273

chronologies show considerable differences in the amplitude and nature of tree-ring response at
these sites. This may be attributed to a number of factors such as site-to-site differences in
microclimate or other site factors, differences in response between species (e.g., the Larch Valley
sites), differences in vigour between sites due to age (e.g., Athadome and Icefield), differences
in standardization procedure and length of record used to derive indices (compare for example,
Larch Valley fir and Athadome spruce). Despite this diversity, a number of common elements
also occur. Except for the two most northerly spruce chronologies, 1799 and 1824 are readily
identifiable (Figures 7 and 8) as conspicuous, narrow marker rings that bracket the period of
particular interest here [in fact, the 1824 ring was missing in 27 of the 35 larch cores measured
at Larch Valley, (Colenutt 1988)]. Generally the records (Figure 7) can be divided into three
parts: a period of relatively high growth, particularly between about 1790 and 1810; a period of
declining growth, usually from about 1810 to 1820 or 1830; and a period of low growth or
general recovery thereafter. The relative intensity, timing and magnitude of the decline varies
between sites but it is particularly marked in the Larch Valley, Lake Louise and Bennington pine
chronologies. At Bennington the oldest pine is not included in this chronology because, following
the sharp decline in ring width in the early nineteenth century, the post- 1820 rings are too narrow
to measure accurately with densitometry.

Several authors, particularly Jacoby and coworkers (Jacoby et al. 1985, 1988; Jacoby and
D'Arrigo 1989; Ivanciu and Jacoby 1988) have reported an abrupt cooling in the early 1800s
based on high-latitude North American tree-ring series and other data. This decline is shown to
some extent by all of the alpine treeline chronologies reported here. At most sites 1816 occurs
in the middle, or at the end of, this period, and is not a marked departure from the trend.
Detailed examination of the index values of these chronologies (Table 2, Figure 8) shows that
only one of the nine chronologies (Larch Valley spruce) has a significantly narrower ring in

1816. Two others have local minimum values in 1816, but these have similar values to preceding
rings in 1814 (Lake Louise) and 1815 (Larch Valley larch). Based on these data, although 1816
is often represented by a narrow ring, the 1814 or 1815 rings are narrower - a fact that cannot
be attributed to the Tambora event. It is not possible, therefore, to detect a marked decline in
growth in 1816 from the ring-width records at these sites.

Filion et al. (1986), Parker (1985), Jacoby et al. (1988) in northern Quebec and Jones et al.
(1988) in Europe report that the 1816 tree ring is distinctive because of its light latewood and low
maximum-density values. Figure 9 shows the available (three) maximum-density indexed
chronologies for the sites previously discussed. Although 1813 appears to be a significant marker
ring, none of these three chronologies show light marker rings associated with 1815, 1816 or

1817. In fact, 1816 appears to have a greater maximum density than adjacent years at these sites
suggesting that, if anything, conditions may have been a little warmer than adjacent years (Parker
and Henoch 1971; Luckman et al. 1985).

Several of the papers in this volume draw attention to the possible effects of the 1783 Laki
eruption and, in preparing this paper, the diagrams were extended to 1780 to include this period.
The data (Figures 7, 8) show considerable variability in the 1780s but Table 3 indicates that 1784
or 1785 is the narrowest ring for the 1780-89 decade in seven of the nine chronologies (Larch
Valley spruce and fir chronologies have slightly lower values in 1782). 1784 is narrower than
1783 in all chronologies and, except for the two northernmost spruce chronologies, this decrease
is marked (7-44%). However, the relative widths of tree rings representing 1784 and 1785 are
inconsistent: at four sites 1785 is much narrower; two sites have 1784 significantly narrower
(including Larch Valley larch, which has a missing ring); and the indexed values are similar at

274

1816

1780 1800 1820 1840 1860 1780 1800 1820 1840 1860

Year Year

Figure 7: Ring-width chronologies (1780-1860) for nine sites in the Canadian Rockies. The ring-width
series are standardized to a mean of 1.0 over the entire period of record (200-600 years;
Table 1) and are plotted at the same scale. The lighter line is the chronology; the thicker line
is a 13-year low-pass filter (see Fritts 1976).

275

1780

Bennington Pine
Bennington Spruce
Robson Spruce

Icefield Spruce

Athadome Spruce

Lake Louise Spruce
Larch Valley Spruce

Larch Valley Fir
Larch Valley Larch

1860

Figure 8: 13-year high-pass filter of ring-width chronologies for nine Rocky Mountain tree-ring sites.

These data are standardized and plotted at the same scale. These data correspond to the
deviations from the low-pass filter curve (Figure 7). 1799 and 1824 are significant narrow
marker rings at most sites. 1783 is the date of eruption of Laki in Iceland.

-io 10

Bennington
Pine

Icefields
Spruce

Lake Louise
Spruce

1780

1800

1820
Year

1840

1860

Figure 9: Standardized maximum density (MXD) chronologies for three sites in the Canadian Rockies.

All are plotted at the same scale. The thin line is the annual indexed value; the thicker line is
a 13-year low-pass filter (Fritts 1976). The shaded year is 1813.

276

ON
00

oo
r-

r-tNcsootni/iin-nO
M OtnMOOOoJ-H

Soo oo ~ m
CI — on On <=> r» CM

CNO-fr — OsON — 00 CN

^cnS\q»noo2;ooTr

r^i/^oortoocot^v^vo
OO— . — on © — 'on —

■a

£0

>
.2P

u
H

o^r-NOt^ocSoocs
— OTj--Hoom§; —

3
u
•o

-o
c

e«

i/5

0£

o
si
B

o
c

4)
CQ

I 1 3

2
"3

J J3

^>
CO

277

the three other sites. Generally, these data indicate that the two years following the 1783 Laki
eruption are considerably narrower, but the pattern is not consistent enough to use 1784 or 1785
as a marker ring (Figure 8).

Conclusions

In this paper we have examined nine ring-width chronologies and three maximum-density
chronologies from four different species at treeline sites in the central Canadian Rocky
Mountains. Although the 1810-20 decade showed a marked decrease in ring-width at all sites,
probably as a result of climatic cooling, there is no indication of a marker (narrow) ring
associated with 1816 or 1817 in the years following the Tambora eruption. Examination of the
three maximum-density series available for these sites indicated no light latewood rings in 1815,
1816 or 1817. It would therefore appear that, unlike sites east of Hudson Bay, the treeline sites
we have examined have no distinctive tree-ring signal to suggest significantly poorer growth
conditions in 1816 or 1817. A major growth decline is identified at all sites during 1810-20 but,
as that decade begins some years prior to the Tambora eruption, that eruption cannot be its
principal cause. This significant period of declining ring-width has been identified elsewhere in
North America, and reflects the most abrupt deterioration in climatic conditions during the last
few centuries.

Acknowledgements

We thank: the Natural Sciences and Engineering Research Council of Canada for support of this
research; Parks Canada and Mount Robson Provincial Park staff for permission to carry out
research at these sites; L. Jozsa, Forintek Canada Corporation, for assistance in the field and in
processing density data; F.F. Dalley (1980), G. Frazer (1981, 1983), S. Ulansky (1982),
J. Hamilton (1983-84), D.C. Luckman (1985-87), D. McCarthy (1986), R. and S. Colenutt
(1987) for coring assistance; and M.I. Johnson, H. Watson, K. Harding, B. Schaus, J. Hamilton
and G. Frazer for ring-width measurements.

References

Baillie, M.G.L. and M.A.R. Munroe. 1988. Irish tree rings, Santorini and volcanic dust veils.
Nature 332:344-346.

Colenutt, M.E. 1988. Dendrochronological studies in Larch Valley, Alberta. B.Sc. thesis,
Geography Department, University of Western Ontario, London, Ontario. 123 p.

Filion, L., S. Payette, L. Gauthier and Y. Boutin. 1986. Light rings in subarctic conifers as a
dendrochronological tool. Quaternary Research 26:272-279.

Fritts, H.C. 1976. Tree Rings and Climate. Academic Press, New York.

Graybill, D.A. 1982. Chronology development and analysis. In: Climate from Tree Rings,
M.K. Hughes, P.M. Kelly, J.R. Pilcher and V.C. LaMarche Jr. (eds.). Cambridge
University Press, Cambridge, pp. 21-28.

278

Hamilton, J. P. 1984. The use of densitometric tree-ring data as proxy for climate at Lake Louise,
Alberta. B.A. thesis, Geography, University of Western Ontario, London, Ontario. 116
pp.

Heusser, C.J. 1956. Postglacial environments in the Canadian Rocky Mountains. Ecological
Monographs 26:263-302.

Ivanciu, I.S. and G.C. Jacoby. 1988. An abrupt climatic cooling in the early 1800s as evidenced
by high-latitude tree-ring data. In: The year without a summer? Climate in 1816. An
International Meeting Sponsored by the National Museum of Natural Sciences, Ottawa,
1986. Abstracts, p. 29.

Jacoby, G.C. and E.R. Cook. 1981. Past temperature information inferred from a 400-year
tree-ring chronology from Yukon Territory, Canada. Arctic and Alpine Research
13:409-418.

Jacoby, G.C. and R. D'Arrigo. 1989. Reconstructed northern hemisphere annual temperature
since 1670 based on high-latitude tree-ring data from North America. Climatic Change
14:39-59.

Jacoby, G.C, E. Cook and L.D. Ulan. 1985. Reconstructed summer degree days in central
Alaska and northwestern Canada since 1524. Quaternary Research 23:18-26

Jacoby, G.C, I.S. Ivanciu and L.D. Ulan. 1988. A 263-year record of summer temperature for
northern Quebec reconstructed from tree-ring data and evidence of a major climatic shift
in the early 1800s. Palaeogeography, Palaeoclimatology, Palaeoecology 64:69-78.

Jones, P.D., K.R. Briffa and T.M.L. Wigley. 1988. Climate over Europe during the summer of
1816. In: The year without a summer? Climate in 1816, An International Meeting
Sponsored by the National Museum of Natural Sciences, Ottawa, 1988. Abstracts, p. 34.

Jozsa, L.A., E. Oguss, P. A. Bramhall and S.G.Johnson. 1983. Studies based on tree ring data.
Report to Canadian Forestry Service, Forintek Canada Corporation, 33 pp.

LaMarche, V.C Jr. and K. Hirschboeck. 1984. Frost rings in trees as records of major volcanic
eruptions. Nature 307:121-126.

Lamb, H.H. 1970. Volcanic dust in the atmosphere: with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society of London
A266:425-533.

Luckman, B.H. 1982. Little Ice Age and oxygen isotope studies in the Middle Canadian Rockies.
Report to Parks Canada, Ottawa. 31 pp.

. 1988. Dating the moraines and recession of Athabasca and Dome glaciers, Alberta. Arctic

and Alpine Research 20:40-54.

279

Luckman, B.H., J. P. Hamilton, L.A. Jozsa and J. Gray. 1985. Proxy climatic data from tree
rings at Lake Louise, Alberta: a preliminary report. Geographie physique et Quaternaire
39:127-140.

Luckman, B.H., L.A. Jozsa and P.J. Murphy. 1984. Living seven-hundred-year-old Picea
Engelmannii and Pinus albicaulis in the Canadian Rockies. Arctic and Alpine Research
16:419-422.

McCarthy, D.P. 1985. Dating Holocene Geomorphic Activity of Selected Landforms in the
Geikie Creek valley, Mount Robson Provincial Park. M.Sc. thesis, University of Western
Ontario, London, Ontario. 304 pp.

Parker, M.L. 1985. Investigating the possibility of a relationship between volcanic eruptions and
tree growth in Canada. In: Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus
55:249-264.

Parker, M.L. and W.E.S. Henoch. 1971. The use of Engelmann spruce latewood density for
dendrochronological purposes. Canadian Journal of Forest Research 1:90-98.

Parker, M.L., L.A. Jozsa, S.G. Johnson and P. A. Bramhall. 1981. Dendrochronological studies
of the coasts of James Bay and Hudson Bay. In: Climatic Change in Canada 2. C.R.
Harington (ed.). Syllogeus 33:129-188.

Watson, H.M. 1983. A dendrochronological study of two sites in Mount Robson Provincial Park.
B.A. thesis, Geography Department, University of Western Ontario, London, Ontario.
147 pp.

280

How Did Treeline White Spruce at Churchill, Manitoba Respond to
Conditions around 1816?

David C.F. Fayle1, Catherine V. Bentley2 and Peter A. Scott3
Abstract

Annual radial increment throughout the stem, and height increment of individual white spruce
trees at Churchill, Manitoba were reconstructed through measurement of ring widths on sections
taken at close intervals throughout the stem. For the period around the eruption of Tambora in
1815, four trees each from open-forest and forest-tundra sites provided data. On each site, one
tree was less than 1 m in height in 1816, the others ranging from 3 to 6 m. Growth of the larger
trees, as indicated by height and radial increment, was generally declining over the two decades
prior to 1816. In the upper stem, particularly of the forest-tundra trees, radial increment was least
in 1818. Effects were less severe in the lower stem and recovery in open-forest trees had begun
in 1818 after a low in 1817. Net-height gain of the forest-tundra trees during 1816-20 was
one-third that of the previous five years, whereas in open forest trees it more than tripled relative
to reduced growth in the previous five years. In combination with the radial-increment data, this
suggests the occurrence of conditions in 1816, or possibly late summer of 1815, that led to
damage of the terminal bud and upper crown with loss of foliage and (or) reduction of foliar
efficiency and production of new foliage. Such effects were much less severe on open-forest
trees. The decline in overall tree growth was statistically significant in 1817-18 compared with
the variability in tree growth for 10 years prior to 1815. Comparisons made with the period
around 1835 (eruption of Coseguina) show subsequent growth reductions were greater than after
Tambora.

Introduction

The relationship between climatic variability and tree-ring widths is often difficult to establish and
unclear at best. However, this relationship can be somewhat clarified by sampling
climate-sensitive trees found in treeline areas where the annual energy deficit is restricting to
growth (e.g., Jacoby and Ulan 1981); inclusion of other tree growth parameters may add
considerable information from the outset (for a review see Fritts 1976).

A severe climatic anomaly that coincides with a large volcanic explosion, such as reported for
the eruption of Tambora during 1815 (Rampino and Self 1982; Catchpole 1985; Parker 1985;
Wilson 1985), offers an opportunity to identify anomalous patterns of tree growth that coincide
with the event (Parker 1985; Filion et al. 1986; Lough and Fritts 1987). By examining
representative samples of tree populations, damage to the forest can be assessed which not only
indicates the climatic impact of such an eruption, but also reveals information on how the climatic
conditions may influence the forest environment.

1 Faculty of Forestry, University of Toronto, Toronto, Ontario M5S 1A1, Canada.

2 R.R. 1, P.O. Box 22, Churchill, Ontario LOL 1K0, Canada.

3 Department of Zoology, University of Toronto, Toronto, Ontario M5S 1A1, Canada.

281

Methods

The field methods and development of the subsequent tree-growth index have been documented
elsewhere (Scott et al. 1988). Briefly, 1 1 white spruce [(Picea glauca) (Moench) Voss)] and two
tamarack [Larix laricina (Du Roi) K. Koch] were harvested in 1982 from five sites near the
treeline at Churchill, Manitoba (58°45'N, 94°04'W). The trees were all open grown and ranged
in age from 88- to 347-years old near their bases. Where identifiable in the upper stem, the
lengths of annual height increments were measured and cross sections cut from their mid-point.
Elsewhere, sections were cut at 10-cm intervals throughout the stem, except where branches were
present. The sections were air dried, sanded and the ring widths measured to 0.01 mm on the
four cardinal directions with a Holman DIGIMIC (Fayle et al. 1983).

Specific volume increment (SVI) was used as a measure of the metabolic activity for a tree in
each year (Shea and Armson 1972). This is the annual volume of wood produced relative to the
surface area of the cambium that produced it (Duff and Nolan 1957); mathematically SVI is the
average width of the growth layer. An advantage of SVI is that it is not a unidimensional
parameter, such as ring width, because it integrates both diameter and height. Furthermore, since
the reference point is a unit area of cambium, a common base is provided for comparison
between trees.

To develop the tree-growth index, the SVI series for each tree was standardized with a robust
estimator (Draper and Smith 1981) using a negative exponential or, in the case of negative
indices, a straight line of negative slope or through the average. The standardized SVIs were
converted to ratios of the individual growth curves and then averaged to produce the final growth
index.

The final growth index was based on all trees sampled. However, only four of the white spruce
from the open forest and four from the forest-tundra were present around 1816. Three from each
type were greater than 3 m in height at that time and were used to analyze radial-longitudinal
patterns of increment in relation to possible influences of climate. The fourth tree from each type
was less than 1 m in height around 1816, and did not provide sufficient information for this
particular purpose.

Results and Discussion

All of the trees show a decline in SVI of varying magnitude either during 1815 or in 1816 which
persists for one to three years following (Figure 1, top and centre). The individual lags in
response and magnitude do not allow for immediate conclusions regarding conditions during the
summer of 1816. However, if we examine the overall status of the regional tree-growth index
10 years prior to 1815, the growth during 1817 and 1818 is below the 95% confidence interval
from what would be expected (Figure 1, bottom). The inference that a volcanic eruption may
influence tree growth is strengthened by repeating the confidence interval test for the 1835
period. The eruption during 1835 of Coseguina, which is much closer to Churchill than Tambora,
may have had more potential for a stronger impact. In fact the 1835 period is the largest sudden
decline in growth throughout the 1710-1982 period of the index.

The cumulative net-height growth patterns for the individual open-forest trees do not indicate any
consistent deleterious effect subsequent to 1815 (Figure 2a). Indeed, height increment appeared
to be slowing down during the previous decade and recovered shortly thereafter (Figure 2b). In

282

contrast, net-height growth was affected in the forest-tundra trees, where it was reduced for
several years before recovering in the 1820s. The greater loss of terminal growth on the
forest-tundra than on the open forest trees is reinforced by the similarity in pattern after 1835
(Figure 2b); a substantial net gain in height did not occur on the forest-tundra trees for two
decades.

The yearly longitudinal distribution of ring width throughout the tree stems for 1815-20 shows
that changes did not occur uniformly (Figure 3). Reductions from 1815 to 1818 were greatest in
the upper rather than lower stem, and more severe on the forest-tundra than on open-forest trees.
The occurrence in the upper part of the growth layers of a 'bulge' in ring width, such as in 1819
for Al and W2, may be the influence of lateral-branch development and (or) an increase in foliar
amounts following damage to the current terminal or existing foliage.

Minimum widths throughout the stem of the average forest-tundra tree occurred in 1818 with a
58%, 50% and 27% reduction compared to 1815 in the upper 0.5 m, upper 0.5-2.0 m, and basal
0.5-2.0 m respectively (Figure 4). In the average open-forest tree, the minimum occurred in 1818
in the upper stem, but recovery was underway in the lower stem with the minimum occurring in
1817; reductions during 1815-18 were 24%, 14% and 3% for the upper 0.5 m, upper 0.5-2.0
m and basal 0. 5-2.0 m respectively.

The reductions in height growth and in ring width in the upper stem subsequent to 1815 and 1835
(Figure 3) indicate damage to, or loss of, the terminal buds and of foliage. The contribution of
photosynthates and growth hormones by a branch to stem growth is related to the amount and
proportion (by age) of the foliage it bears, the distance of this foliage from the stem and the
amount of light it receives. A relatively short branch system with a high proportion of well-lit
young foliage will make a high contribution to stem growth.

In white spruce, the number of new needles that will be produced in the current year, and the
potential shoot elongation were determined when the bud was formed in the late summer of the
previous year (Owens et al. 1977). The degree of elongation and production of photosynthate are
determined by conditions in the current year. Needles can be retained for 10-15 years at Churchill
but there is a loss of photosynthetic efficiency with age. Current, one- and two-year- old foliage
may contribute as much as 60% of the total assimilation in white spruce (Clark 1961). The
influence of a favourable or unfavourable part of or whole growing season will therefore not only
have different, but also lag effects, on growth, which can be compounded if there is a physical
loss of new needles or premature loss of old needles.

We have found that the loss of the terminal bud or shoot in treeline white spruce at Churchill is
a common phenomenon often occurring at the same time as reduced width of the growth layer
throughout the tree, but particularly in the upper stem. Poor growth occurs for several years
while a lateral bud or branch establishes itself as the new terminal. Loss can come about through
direct or indirect causes. An example of the latter could occur if the roots remain frozen while
growth is under way (e.g., Scott et al. 1987), leading to desiccation and death of the needles and
buds in the entire upper part of the tree (Sakai 1970; Kullman 1988). The proportion affected will
determine the degree of growth reduction and, in combination with ongoing climatic conditions,
the length of time to full recovery. The difference between open-forest and forest-tundra trees
may be in the more exposed nature of the latter and the longer retention of older needles on the
former, which provides a greater reserve. It is clear, from the slow recovery after the decline in
growth during 1835, that many trees were apparently damaged this way, although there is little
evidence of this occurring subsequent to 1815.

283

FOREST -TUNDRA OPEN FOREST

0.4 -1

YEAR

Figure 1: The specific volume increments (SVI) of the four open-forest and forest-tundra white spruce
(top) that compose the growth index during the period around 1816. (This index, shown in the
centre, is based on the SVIs of 13 trees.) Enlargement of the 1805-45 years (bottom) includes
the mean (dashed line) and upper and lower (dotted line) 95% confidence limits for the 10-year
periods prior to the 1815 eruption of Tambora and the 1835 eruption of Coseguina, to show that
the years following these eruptions exhibit unusually poor growth.

284

HEIGHT

(CM)

r50

410
530
565

505

250

350
505

370

Figure 2a: Cumulative-height
curves for the three
open-forest (top) and
forest-tundra (bottom)
trees and their
respective averages
(heavy line) during
1800-30. The numbers
at the left give the
total height of the
trees in 1800, to the
nearest 3 cm. The
circles and squares
indicate section
heights from which
the curves were
constructed. The
vertical line marks
1815.

1800

YEAR

(J 0J

ID
I

1 I I I I I I I I I I 1

1801-5 6-10 11-15 16-20 21-25 26-30 31 35 36 4041-45 46-50 51 55 56 60 61-65 66-70

FIVE-YEAR INTERVAL

Figure 2b: The net-height increase during five-year periods for 1 801-70 inclusive for open-forest (hatched
bar) and forest-tundra (open bar) trees. The five-year periods immediately following the
eruptions of Tambora and Coseguina are underlined.

285

2-\
1

1815 16 17 18 19

1815 16 17 18 19 20

1835 36 37 38 39

YEAR

S W2

1835 36 37 38 39 40

Figure 3: The width of the growth layer (average of four radii) for the three open-forest (OF) and three
forest-tundra (FT) trees during 1815-20, and for one example of each for 1835-40.

286

1 -I

I .
I

."/

5 -

\ •
i

\ •

• V

TOP 0 - 50 CM

\ A /V

t / v

/
i
/
/

i '

I 1 1

BASAL 50-200CM

1800

YEAR

Figure 4: The average ring width for the 0-0.5 m and 0.5-2.0 m intervals from the contemporary apex,
and for the 0.5 - 2.0 m interval above the stem base, for open-forest (dashed line) and
forest-tundra (dotted line) trees. The vertical line marks 1815.

287

The fact that radial growth is least in 1818 rather than 1816 is probably the result of cumulative
effects arising from adverse conditions in 1816 or late summer of 1815. We suggest that the
occurrence of conditions at that time led to damage of the terminal bud and upper crown with loss
of foliage and (or) overall reduction of foliar efficiency and production of new foliage. Adverse
conditions in August, after height increment was completed in either or both years, would reduce
the amount of foliage produced the following year. Recovery of the open-forest trees below the
apical 0.5 m in 1817, and a lesser reduction from 1817 to 1818 than from 1816 to 1817 in the
forest-tundra trees, suggests that growing conditions were improving in late 1817. The reduced
growth of the forest-tundra trees in 1818 may have been due to the cumulative effects of reduced
foliar area, particularly in younger age classes, rather than adverse growing conditions per se.
All trees showed improved radial growth in 1819 suggesting favourable conditions for bud
development existed in 1818 (Figure 5).

-4-,

a. A) ±

YEAR

Figure 5: Diagrammatic presentation of reduction in shoot growth due to adverse effects on bud
development, shoot and needle elongation without (left) and with (right) damage to the terminal
bud, shoot and needles. Horizontal lines represent needles of different age classes, from current
year (heavy line) to five-years old (dotted line). Older years are not shown. The brackets
indicate the needle classes that would normally contribute the bulk of photosynthate.

Year one (e.g., 1815) shows normal growth and bud development. In year two unfavourable
conditions throughout the growing season restrict shoot and needle elongation and bud
formation. Incipient damage to the bud and needles may occur. In year three, growing
conditions are more normal and needle elongation and bud formation are not restricted, but the
amount of new foliage is reduced due to previous adverse conditions. Where damage occurred,
a lateral bud may begin to assume dominance, but its small size has restricted the amount of
needles produced. In year four, growth and development are near normal. Where damage had
occurred, the quantity of photosynthetically-efficient foliage is still low and ring width is
minimal. Recovery occurs in year five here, whereas it was already underway in the undamaged
shoot.

288

The above scenario is complemented by the observations of Filion et al. (1986) who reported a
high occurrence of 'light rings' in 1816 and 1817 in krumholz black spruce in northern Quebec.
We have not had the opportunity yet to determine their presence in our trees. If they do occur,
which is likely, unfavourable conditions in the late part of the growing season and (or) a shortage
of photosynthate, the result for example of needle loss, are suggested. In the 'light rings'
illustrated by Filion et al. (1986), the last formed tracheids in the annual ring show a normal,
narrow radial diameter but wall thickening is minimal. A supply of photosynthate is required to
complete the process of wall thickening and environmental conditions must permit the completion
of the maturation process for normal latewood formation.

From the examination of tree growth during 1815 and particularly 1835, it appears that a
stochastic event, such as a volcanic eruption, occurring many thousands of kilometres away may
have a widespread detrimental effect on forest productivity. Climatic conditions at Churchill
following the eruption of Tambora in 1815 and Coseguina in 1835 did have adverse effects on
growth of white spruce. Correspondingly, Wilson (1985) reports that, on the east side of Hudson
Bay, conditions were poor during the late summer of 1815, during 1816 and possibly the first
half of 1817. Similarly, while it is apparent that 1816 was not truly without a summer at
Churchill, it may have been one of 5°C temperatures instead of the long-term average of 10°C.

Acknowledgements

We thank Ed Cook and Gordon Jacoby for helpful advice and supplying some analysis programs.
Ring-width measurements were made using facilities of the Ontario Tree Improvement and Forest
Biomass Institute, Ontario Ministry of Natural Resources, with grants supplied to the authors by
Environment Canada and to P. Scott by Indian and Northern Affairs Canada. NSERC provided
travel funds for D. Fayle. We also thank Roger Hansell for his help in the project and C.R.
Harington for his support.

References

Catchpole, A.J.W. 1985. Evidence from Hudson Bay region of severe cold in the summer
of 1816. In: Climatic Change in Canada 5. C.R. Harington (ed.). Syllogeus 55:121-146.

Clark, J. 1961. Photosynthesis and respiration in white spruce and balsam fir. Syracuse
University, State University College of Forestry Technical Publication 85:1-72.

Draper, N. and H. Smith. 1981. Applied Regression Analysis. Second Edition. John Wiley
& Sons, Inc., Toronto. 709 pp.

Duff, G.H. and N.J. Nolan. 1957. Growth and morphogenesis in the Canadian forest
species. II. Specific increments and their relation to the quantity and activity of growth in
Pinus resinosa Ait. Canadian Journal of Botany 35:527-572.

Fayle, D.C.F., D.C. Maclver and C.V. Bentley. 1983. Computer-graphing of annual ring
widths during measurement. The Forestry Chronicle 59:291-293.

Filion, L., S. Payette, L. Gauthier and Y. Boutin. 1986. Light rings in subarctic conifers as a
dendrochronological tool. Quaternary Research 26:272-279 .

289

Fritts, H.C. 1976. Tree Rings and Climate. Academic Press Inc. New York, New York.
567 pp.

Jacoby, G.C. and L.D. Ulan. 1981. Review of dendroclimatology in the forest-tundra
ecotone in Alaska and Canada. In: Climatic Change in Canada 2. C.R. Harington (ed.).
Syllogeus 33:97-128.

Kullman, L. 1988. Subalpine Picea abies decline in the Swedish Scandes. Mountain
Research and Development 8:33-42.

Lough, J.M. and H.C. Fritts. 1987. An assessment of the possible effects of volcanic
eruptions on North American climate using tree-ring data, 1602 to 1900 A.D. Climatic
Change 10:219-239.

Owens, J.N., M. Molder and H. Langer. 1977. Bud development in Picea glauca. I.
Annual growth cycle of vegetative buds and shoot elongation as they relate to date and
temperature sums. Canadian Journal of Botany 55:2728-2745.

Parker, M.L. 1985. Investigating the possibility of a relationship between volcanic eruptions
and tree growth in Canada (1800-1899). In: Climatic Change in Canada 5.
C.R. Harington (ed.). Syllogeus 55:249-264.

Rampino, M.R. and S. Self. 1982. Historic eruptions of Tambora (1815), Krakatau (1883),
and Agung (1963), their stratospheric aerosols, and climatic impact. Quaternary Research
18:127-143.

Sakai, A. 1970. Mechanism of desiccation damage of conifers wintering in soil-frozen areas.
Ecology 51:657-664.

Scott, P. A., C.V. Bentley, D.C.F. Fayle and R.I.C. Hansell. 1987. Crown forms and
shoot elongation of white spruce at the treeline, Churchill, Manitoba, Canada. Arctic and
Alpine Research 1 9 : 1 75- 1 86.

Scott, P. A., D.C.F. Fayle, C.V. Bentley and R.I.C. Hamsell. 1988. Large scale changes in
atmospheric circulation interpreted from patterns of tree growth at Churchill, Manitoba,
Canada. Arctic and Alpine Research 20: 199-21 1 .

Shea, S.R. and K.A. Armson. 1972. Stem analysis of jack pine (Pinus banksiana Lamb.):
techniques and concepts. Canadian Journal of Forest Research 2:392-406.

Wilson, C. 1985. The Little Ice Age on Eastern Hudson/James Bay: the summer weather
and climate at Great Whale, Fort George and Eastmain, 1814 to 1821, as derived from
Hudson's Bay Company records. In: Climatic Change in Canada 5. C.R. Harington (ed.).
Syllogeus 55:147-190.

290

The Climate of Central Canada and Southwestern Europe Reconstructed
by Combining Various Types of Proxy Data: a Detailed Analysis of the
1810-20 Period

J. Guiot1

Abstract

In this study, an attempt is made to synthesize various kind of proxy records and to reconstruct
complete climatic series. In central Canada, tree-ring series and historical records from the
Hudson's Bay Company (i.e., ice-condition and early instrumental data) have been assembled to
reconstruct a seasonal temperature and sea-level pressure network back to 1700. The summer of
1816 was among the coldest in the period studied. The beginning of the nineteenth century was
also cold, especially after 1807, but the main characteristic is great variability (e.g., 1818 was
one of the warmest years since 1700). These results are compared with those obtained in a similar
manner for southwestern Europe and northwestern Africa on the basis of tree-ring series, 180
records, wine-harvest and other archival data. The 1810-20 period was also among the coldest
of the last millennium, and 1816 was one of the four coldest years since the eleventh century. In
the Mediterranean region, this period was far less cold. Some details are given on the method
used for these reconstructions. As the proxy series are not homogeneous, particular devices are
needed to estimate missing data and to reconstruct low-frequency components. The techniques
are adapted from multiple regression, digital filtering, bootstrap analysis and principal component
analysis.

The Data in Central Canada

The meteorological network is made up of 67 stations selected from the meteorological database
of the Atmospheric Environment Service of Canada. The period of analysis is restricted to
1925-83 and the region is delimited by 61° - 105°W and 47° - 73°N. The monthly data are
averaged into seasonal series.

The second Canadian data set is built from the proxy series (Figure 1) available at the date of the
study (Guiot 1985a), including:

• freeze-up and break-up dates of rivers entering the western shore of Hudson and James bays:
nine series (1714-1871 at maximum) derived by Catchpole and Moodie (1975) and extended
to the modern period using recent data (Allen 1977);

• freeze-up and break-up dates of the Red River at Winnipeg, extending from 1798 to 1981: two
series build by Rannie (1983);

• monthly temperature data for York Factory (1774-1910) and Churchill Factory (1768-69/
1811-58) derived by Ball and Kingsley (1984), spatially and temporally averaged into four
seasonal series for the York-Churchill region.

CNRS UA 1152, Laboratoire de Botanique Historique et Palynologie, Faculte de St. Jerome, 13397 Marseille
Cedex 13, France.

291

These data are completed by tree-ring series. The trends of these series are modeled by negative
exponentials, polynomials or filtered curves by the authors as proposed by Fritts (1976). Indexed
series are obtained by dividing each ring width by its trend. The series used are the following:

• two white spruce ring-width indices series from Nain, Labrador (1769-1973) and Border
Beacon, Labrador (1660-1976) from Cropper and Fritts (1981);

• one larch ring-width indices series from Fort Chimo (now Kuujjuaq), Quebec (1650-1974) also
from Cropper and Fritts (1981);

• two white spruce series from Cri Lake, near Kuujjuarapik, Quebec, (1750-1979) (Parker et
al. 1981), the first being ring-width indices and the second being ring maximum densities;

• two white spruce ring-width indices from Churchill, Manitoba (1691-1982) by P. Scott in
Hansell (1984), the first was sampled in open forest and the second in forest-tundra.

Finally a total of 22 proxy series are available to reconstruct temperature in central Canada for
1700-1979.

Figure 1: Location of the proxy-series sites in Canada.
The Data in Europe and North Africa

Meteorological data are the annual series gridded by Jones et al. (1985) extending from 35° to
55°N by steps of 5°, and from 10°W to 20°E by steps of 10°. So 20 series are available from
1851 to 1984, with missing data mainly before 1900.

292

The second data set, the proxy series, are collected from the longest proxy series existing for
Europe and Morocco (Figure 2). A part of them consists of tree-ring chronologies of various
species from various sites. They are also detrended as suggested by Fritts (1976). The set
includes:

• oak ring-width series from west of the Rhine, near Trier, Germany (820 to 1964) collected
and indexed by Hollstein (1965);

• oak ring-width series from the Spessart forest area (50°N, 9°30'E) in Germany (840 to 1949)
collected and indexed by Huber and Giertz-Siebenlist (1969);

• oak ring-width series from Belfast, Northern Ireland (1001 to 1970) collected and indexed by
Baillie (1977);

• oak ring-width series from southwestern Scotland (946 to 1975) collected and indexed by
Pilcher and Baillie (1980);

• pine ring-width series from southern Italy (1148 to 1974) collected and indexed by
Serre-Bachet (1985);

• larch ring-width series from Vallee des Merveilles, southern French Alps (1100 to 1974)
collected and indexed by Serre (1978);

• fir ring-width series from Mont Ventoux, southern France (1660 to 1975) collected and
indexed by Serre-Bachet (1986);

• pine ring-width series from northern Italy (925 to 1984) collected and indexed by Bebber
(personal communication);

• larch ring-width series from Orgere, northern French Alps (1353 to 1973) collected and
indexed by Tessier (1981);

• two larch ring-width series from Mercantour, southern French Alps (1701 to 1980 and 1732
to 1981) collected by Guibal (personal communication) and indexed for this study.

Another group of proxy series is composed of data derived from archives. These historical data
have been compiled by various historians and/or climatologists:

• decadal temperature estimates of Bergthorsson (1969) for Iceland (1050 to 1550). These data
were analyzed by Ogilvie (in Ingram et al. 1978), and those before 1 170 and after 1450 were
reported as unreliable - the unreliable decades are considered as missing;

• summer temperature index of Bray (1982) based on German and French wine-harvest data and
central England (Manley) temperatures (1453 to 1973);

• the Pfister (1981) thermal indices in Switzerland, averaged on an annual basis from 1550 to
1829;

293

• Three glaciological series (180)
• • Greenland

• Iceland

17 Tree-ring series
(Morroco)

Figure 2: Location of the proxy-series sites in Europe and Morocco.

• the mean annual dates at the beginning of the grape harvest in northeastern France, French
Switzerland, and southern Rhineland of Le Roy Ladurie and Baulant (1981) (1484 to 1879);

• the average annual dates at the beginning of the grape harvest in Switzerland reported by
Legrand (1979) (1502 to 1979);

• frequency of southwesterly surface winds in England (1340-1978) from direct observations
(1669 to 1978) in the London area and from historical proxy data before. These data are
reconstructed by Lamb (1982);

A last category is provided by 180 data in the Arctic ice. These isotopic series can be considered
as good indicators of temperature, since the condensed vapour is enriched in heavy isotopes:

• Camp Century, Greenland, 180 quasi-decadal values (1200 to 1970) collected and analyzed by
Dansgaard et al. (1971);

• two isotopic series in central Greenland, 30-year running means of annual maxima of 180 in
ice cores compiled by Williams and Wigley (1983) (1180 to 1800).

294

Finally, to these data are added the first three principal components of the 17 longest cedar
ring-width series in Morocco, sampled by A. Munaut and C. Till and analyzed by Till (1985).
These series (1068 to 1979) represent nearly 40% of the total variance of the 17 raw series.

The period 1068-1979 is retained for a total of 23 series. To simplify matters, European data will
include both European and northwestern African data.

Data Conditioning

The predictand matrix as well as the predictor contains missing data. Therefore it is fairly natural
to estimate the gaps before beginning any detailed analysis. For the Canadian data, the method
employed to estimate the missing data is explained in Guiot (1985a) and, with more details, in
Guiot (1986). The general procedure is similar to that used for the management of the European
series, described here.

The best analogues method is used to estimate the missing data of the proxy-series matrix. The
main advantage of this method is that we have not to assume any linear relationship between the
variables. This is particularly recommended when, like here, the series are highly heterogeneous.
The estimate of a missing observation for a given series is provided by the most similar
observations (analogues) of the same series within the 1200-1900 interval, the distance between
observations being established on the m^ observations of the series available.

mik

dk2= E(Xij-Xkj)2 (i)
j = 1

The observations, denoted by k, available among the 20 best-fit analogues of observation i
provide the wanted estimate

5^ k Xui/diu2

x, = — (2)

Ekd*2

The correlation between estimates and actual values computed on the available data and averaged
on the 23 series is 0.73 (ranging from 0.45 to 0.86), which is highly significant. For observations
outside the 1200-1900 period, the mean correlation remains high, say 0.60 (ranging from 0.22
to 0.89). It must be noted that we have not estimated any coefficients so that the statistics
computed on the calibration interval as well as on the verification one can be considered as
independent. The mean and the standard deviations of the estimates are quite close to the actual
ones, with discrepancies less than 15% of the mean standard deviation. Depending on the number
of degrees of freedom, we can consider that the estimates are reliable.

For the meteorological data matrix, multiple regression was used. This method cannot be applied
directly because the number of regressors is not constant on the total calibration interval (1851-
1984). If m; is the number of regressors available for observation i (i.e., with no missing data),
the regression equation may be written as follows:

295

X;j = aoj + 52 akj x* (3)
k = 1

The correlation between estimates and actual values averages 0.76, with the highest values in the
Northwest (0.90). These coefficients are highly significant, but the estimates must be considered
as less reliable at the southern margin of the region analyzed (correlation around 0.70).

Extrapolation of the Temperature Series

When the predictor and predictand matrices, are fully determined, it is possible to extrapolate the
annual -temperature series from the proxy series, using the common observations to calibrate a
relationship. It is advisable first to transform the raw series into principal components (PCs).
Indeed, a large proportion of the high order PCs represents extremely small proportions of
variance, so that they can be assumed to be indistinguishable from statistical noise.

Reduction of the Number of Variables

For the European annual-temperature series, 10 PCs are used explaining together around 90%
of the variance. For the European proxy series, 19 PCs are used explaining 95% of the variance.
For the Canadian season temperature series, the first four PCs used explain between 82.5% (for
summer) and 91% (for autumn) of the total variance. For the proxy series, the number of PCs
depends on the season reconstructed.

Bootstrap Regression

In central Canada, a multiple regression has been employed to calibrate the relationship between
climate and proxy series. In Europe, a more sophisticated approach, termed bootstrap regression,
seemed more advantageous.

Bootstrapping is a recent technique devised by Efron (1979) to estimate statistics for unknown
population distributions by Monte Carlo simulations. The idea is to resample the original
observations in a suitable way to construct pseudo-data sets on which the estimates are made. In
regression, this is particularly useful when the residuals are non-normal or autocorrelated, or
when the data set is too small.

The bootstrap method is in fact a generalization of jackknife replication. The frame of the method
can be summarized in a few lines. From the interval (l,n), where n is the size of the original data
set, n pseudorandom numbers are randomly taken with replacement using a uniform distribution
protocol. These n numbers are used to resample the actual observations. We should insist here
on the fact that an observation is the vector of the m proxy data and p climatic parameters
corresponding to the same year. The n observations selected in this way provide a pseudo-data
set. This is repeated an arbitrary number NC times, and at each time, a regression is computed.

The reliability of a particular statistical model must be assessed by calculating a number of
verification statistics measuring the degree of similarity between predictand observations and their
estimates for time periods independent of the calibration. So a successful reconstruction is one
for which it is demonstrated that independent estimates continue to be accurate at a level greater
than would be expected solely by chance. "The process used to optimize the coefficients of the
model virtually ensures that the results will be more accurate for the calibration data than for any
other observations to which it may be applied. It is why the decreasing of accuracy should be

296

measured whenever possible. " (Fritts and Guiot 1988). Bootstrap regression enables one to
integrate verification in the calibration process and to use the n observations both for the
calibration and the independent verification:

• for each of the NC replications, the regression coefficients are computed and applied to proxy
series to obtain the corresponding reconstruction;

• the reconstruction is compared to the actual climatic series both on the set of retained
observations and on the others; thus verification statistics are calculated NC times;

• the mean and standard deviations of the verification statistics are obtained on the dependent
and independent data set;

• the final reconstruction is given by the median of the NC replicated reconstructions, and a
90% -confidence interval is given by the 5th and 95th percentile.

Decomposition of the Spectra into Two Bands

Before computing a bootstrap regression, the predictors and the predictands are filtered, so that
their spectra are decomposed into two bands (Guiot 1985b). Once more, the method is illustrated
with European data. We use a nine-weights low-pass filter with a cut-off period of seven years.
The effect of this filter is illustrated in Figure 3 with the first PC of the proxy series. The
complementary high-pass filter enables us to retain the short-term fluctuations of the series. The
raw series is the sum of both low-frequency and high-frequency components (Figure 3).

In the two frequency bands, bootstrap regressions are calibrated on the common period,
1851-1979. This method is particularly necessary for the low-frequency components dominated
by large autocorrelations, which induce troubles in the interpretation of the fit quality. The
"abnormality" of these smoothed data is compensated for by a lot of simulations.

Table 1 presents some statistics useful for the evaluation of the regressions. For each of the 50
simulations, the estimated means and standard deviations are compared to the actual ones on the
randomly-drawn observations, as well as on the others. The deviations of these statistics are
averaged over the 50 simulations (Table 1). Apparently the standard deviations are slightly
underestimated, as expected, and the biases are not greater on the independent observations.
Concerning the calibration data set, the correlations between estimated and actual observations
are lower for the high-frequency components than for the low-frequency components.
Nevertheless this must be appreciated regarding the reduced number of degrees of freedom of
autocorrelated series. The most important feature is the lack of stability, appearing in the
independent data set, of the high frequencies for components 3 to 6 and 8 to 10, while the low-
frequency components estimates are quite stable. This justifies the spectral decomposition.

The regression coefficients are applied in each band to extrapolate the 10 temperature PC series
back to 1068. The entire spectra are recomposed by adding the reconstructed high-frequency PC
series to the low-frequency ones. Table 2 presents the effect of this addition for two periods in
the calibration period. During the first period (1851-1900), temperature observations are less
abundant and of lesser quality than during the second (1901-79). The means and standard
deviations in the older period appear to be systematically more underestimated than in the more
recent one by a factor of 2. These underestimates are negligible for the means: they represent
between 0.7 and 2.5% of the total variance of the PCs (that is 2.104). The biases are higher for

297

I860 1880 1900 1820 i8<i0 1880

Figure 3: The first principal component of the European proxy series (1850-1979) and its spectral
decomposition.

298

Table 1: Verification Statistics for the Reconstruction of the Low- and High-Frequencies Component
of the First 10 PCs of the Annual Temperatures. [These statistics are averaged on the 50
replications: (a) on the calibration observations (randomly drawn); (b) on the others. dM =
estimated mean minus actual mean; dS = estimated standard deviation minus actual standard
deviation; R = correlation coefficient between estimates and actual variables ± 1 standard
deviation.]

Low Frequencies

Var.

(a)

0.00

(b)

-0.01

0.01

-0.01

0.00

0.04

0.01

-0.01

(a)

-0.15

-0.28

-0.30

-0.13

-0.33

-0.21

-0.17

-0.22

-0.34

-0.17

(b)

-0.15

-0.30

-0.29

-0.11

-0.33

-0.22

-0.17

-0.22

-0.33

-0.19

(a)

0.85

0.72

0.70

0.87

0.67

0.79

0.83

0.78

0.66

0.83

± 0.03

0.03

0.02

0.04

0.03

0.04

0.03

(b)

0.74

0.53

0.57

0.83

0.48

0.65

0.75

0.67

0.47

0.75

± 0.09

0.09

0.04

0.10

0.09

0.05

0.07

0.09

0.07

High Frequencies

Var.

(a)

0.00

(b)

0.02

-0.01

0.04

-0.01

0.01

0.03

0.00

-0.01

0.03

(a)

-0.35

-0.50

-0.44

-0.54

-0.49

-0.62

-0.42

-0.48

-0.57

-0.51

(b)

-0.29

-0.45

-0.41

-0.54

-0.43

-0.56

-0.40

-0.43

-0.53

-0.49

(a)

0.65

0.50

0.56

0.46

0.51

0.38

0.58

0.52

0.43

0.49

± 0.04

0.05

0.06

0.04

0.06

0.04

0.07

0.05

(b)

0.48

0.24

0.18

0.12

0.19

-0.10

0.31

0.15

-0.01

0.11

± 0.09

0.09

0.11

0.15

0.10

0.11

0.09

0.11

the variability: between 10 and 20% of the total variance. These results are clearly better than
those obtained without spectral decomposition. With this last method, the correlation gain is about
0.14, and the underestimating factor is divided by 2. We conclude that spectral decomposition
increases the quality of fit, but we cannot infer that this best fit is warranted on independent
periods. The example dealt with in Fritts and Guiot (1988) nevertheless confirms the stability of
such extrapolations.

The reconstruction of the raw-temperature series is obtained by postmultiplying the 10 PC series
matrix by the eigenvector matrix. The mean correlation between estimated and actual values is
0.63 (from 0.46 to 0.76) on the 1851-1979 period, with a maximum of more than 0.70 in the
Northwest. These reconstructed series are provided with a 0.90-level confidence interval that is
0.3 °C in mean.

Analysis of the Reconstructions as a Whole

The reconstructed series in Europe are analyzed from different points of view. The trend of the
16 series, at latitudes ranging from 40° to 55 °N is plotted in Figure 4, on the basis of 20-year

299

Table 2: Verification of the Reconstructions of the 20 Annual -Temperature Series PCs (Multiplied
by 100). The sum of the low and high frequencies are first verified, and then the regression,
without separating low from high frequencies: (a) actual means; (b) estimated mean; (c)
actual standard deviations; (d) estimated standard deviations; (e) correlation between actual
and reconstructed PCs. SSD = sum of squared differences [sign (-) means underestimates].

Low and High Frequencies

1901-

1979

1851

-1900

PC01

-41

270

187

0.69

296

195

0.73

PC02

167

0.61

-55

170

0.51

PC03

-17

-19

154

0.50

190

0.61

PC04

-15

-18

100

0.83

104

0.57

PC05

-11

0.37

-24

. 148

0.79

PC06

-34

-28

0.47

0.59

PC07

0.68

-3

-1

0.74

PC08

0.62

-16

0.71

PC09

-7

0.58

-3

-6

0.47

PC 10

-8

-5

0.56

0.72

SSD

(-)1504

(-)20867

(-)5290

(-)43559

mean

0.59

0.64

Standard Regression
1901-1979 1851-1900

PC01

-41

270

144

.61

296

160

.65

PC02

-1

167

.35

-55

-1

170

.23

PC03

-17

-21

154

.37

190

.58

PC04

-15

-7

100

.74

104

.45

PC05

-5

.27

-24

148

.62

PC06

-34

-2

.33

.41

PC07

.62

-3

.58

PC08

.47

-16

-14

.51

PC09

-7

-3

.38

-3

-9

.43

PC 10

-8

-1

.27

-4

.60

SSD

(-)4498

(-)53549

(-) 11393

(-)83891

mean

.45

.51

periods. A climate generally colder than now, results interrupted by a few warm periods. A
spatial distribution of the anomalies is presented at some key 20-year periods for the 20
gridpoints.

Before 1200, the temperature was extremely low in the whole region. From 1200 to 1400, the
analyzed region experienced a relatively warm climate mainly in the Southwest. This warm
period is often called the "Little Climatic Optimum" of the Middle Ages. The results for the
1070-1420 period are verified by comparing the indices of Alexandre (1987) that are valuable for
Mediterranean and non-Mediterranean western Europe. In order to make these indices
representative of both winter and summer, we have used the difference "Winter severity index
minus Summer precipitation index" to represent the annual temperature. It confirms that a

300

warming is obvious from the beginning of the thirteenth century, with a mean of -1 .8 before 1220
and 0.23 after (Figure 4).

From 1420 to 1460, conditions were very cold, except in North Africa. The warming at the end
of the Middle Ages lasted from 1460 to about 1550. The spatial distribution was similar to that
of 1200-1400, with a maximum to the West.

The period generally called the "Little Ice Age" seems to have begun about 1500. The first part
of this period (1550-1610) was effectively cold mainly in areas along a diagonal extending from
the British Isles to Tunisia. The seventeenth century was generally warm in the Southwest. In the
Southeast, the cooling started around 1550. In fact, the Little Ice Age really started at the end
of the seventeenth century. Temperature was low everywhere except in the Southwest. It lasted
until 1860, with two particularly cold periods about 1700 and 1815. It had no equivalent in the
Southwest, although the precipitation reconstructions of Till and Guiot (1988) indicate increasing
moisture - particularly during these two extreme episodes. Richter (1988) confirms the climatic
differences between the southwestern Mediterranean Basin and northwestern Europe. Its
reconstruction of summer precipitation from pine tree-ring series shows that 1810-20 was wet in
central Spain and its reconstruction of winter temperature shows that the same area was warm.
As central Spain is located at the midpoint between Morocco and the rest of western Europe,
these reconstructions are simultaneously, a confirmation of our temperature reconstructions and
the precipitation reconstructions of Till and Guiot (1988).

The modern warm period began in the mid-nineteenth century, with a maximum between 1930
and 1950 - especially in the Northwest and in the Southeast.

Similar reconstructions for the four seasons have been obtained in central Canada, but only for
the last three centuries. The reconstructions are detailed in Guiot (1985a). WI focus here on a
comparison with Europe. Figure 5 shows three synchronous long periods on both continents:
1700-50; 1780-1820; and 1850-1920. After the beginning of the twentieth century, the general
warming appeared in Canada some five years later than in Europe. The synchronism during the
Little Ice Age could mean that it is forced by an external common phenomenon.

The Year 1816

If 1810-20 appears, as a whole, very cold in Europe, it is highly variable in Canada (Figure 5).
For example, the summer was very cold in 1816 but it was very hot two years later. On the two
continents, the cooling began at the beginning of the nineteenth century. 1816 is only a period
where this cooling reached an extreme. If the volcanic eruption of Tambora in 1815 had an
influence on the severe climate of the following year, it only accentuated a trend, and its eventual
effects must be placed in the context of the cold period of the "Little Ice Age". This trend is
noticeable as well in Canada as in Europe. Figure 5 also shows that the summer of 1816 was the
coldest of the last three centuries in central Canada. In Europe, we have found three other years
as cold as 1816: 1081, 1454 and 1703. The four coldest years of the millennium reached mean
anomalies of -1.5°C (in the range 10°W - 20°E and 35° - 55°N).

Central Canada reconstructions are sufficiently precise for a chronology of the cooling in the
region to be established. Figure 6 presents the distribution of temperature anomalies in relation
to 1950-79. The temperature of the North is a blank because the proxy series used is not
representative of latitudes higher than 65°N. Winter 1816 was nearly normal in the whole region
studied except in the Southwest where the cooling was already perceptible (negative anomalies

301

ANNiim TFjIPFk^TI IRF -1. -0,8 -0,6 -0,4 -0,?

I II II iUI IJUI A All IA AJJ II I A UJIU

(anomalies)

EUROPE

; ; W&£fr

jViViViS

COLD

-2-10 1
Alexandre Index

WARM

Figure 4: The 20-year trend of temperature variations in Europe (area restricted to latitudes 40°-55°N
and longitudes 10°W-20°E). The distribution of the anomalies for some characteristic periods
is shown for the total area (including 35°N). The broken horizontal lines represent the
90% -confidence intervals computed by bootstrapping. Between 1070 and 1410, are the
smoothed Alexandre (1987) indices representing winter severity and summer precipitation.

302

CANADA : GIJUULM TEMP

20 31)

■ ■ J i\
' -ri . . .
■ v ■

■A

- ,\ • ,
.i >

r- rl

t ! I.

I : —

I." '

' 'i i

! I

1 •*

■ i'
, i1 ■

CANADA ANNUA! f f UP

30 30

CiW Llinni'E: ANNUAL IE Ml'.

Ti7

/• - - - -

v-=«

... '->

17UO

M !

i:. ■

ll'A

■

;:V -

'i! ■

I. I,

* . I1

r . lj. -'

t Li

1 ■ 1

•'! ■ 1

~T7"

c *l

l'i '

■

Is 1

. I!

■

'■J

— - ,'h ■

I;-

ThvTT-

- ' >

■■\ ■

J_l_

Figure 5: The annual temperature in southwestern Europe and central Canada (both being averages of the
individual reconstructed series) and the summer temperature in central Canada. The series are
smoothed with a digital filter (cut-off period = seven years).

303

Figure 6: The distribution of temperature anomalies for the four seasons in central Canada.

greater than 1°C). The cooling affected the whole region in spring, with a maximum in the
Southwest where negative anomalies of 2°C are reached. In summer, the temperature anomalies
were globally -1°C with a minimum of almost -3°C in the region of Kuujjuarapik (southeastern
Hudson Bay). A secondary minimum of -2°C occurs in the Churchill region (western Hudson
Bay). The Southwest has already begun to warm, since spring anomalies are -2°C and summer
ones -1°C. Autumn is nearly normal everywhere except in the Southwest where the positive
anomalies are + 1°C.

In Europe, because emphasis was laid on the ability to reconstruct temperature series over a
millennium, it is impossible to collect a sufficient number of series to obtain a seasonal
resolution. Figure 7 is nevertheless instructive respecting the spatial differences of the cooling.
Briffa et al. (1988; this volume) have already shown that the summer of 1816 in central Europe
was less cold than in western Europe. This is confirmed as far as the annual temperatures are
concerned, and it is possible to more precisely judge the temperature of the western
Mediterranean Basin. The maximum negative anomalies concern the British Isles and northern
France (-3°C), but they extend far away to Africa - especially Tunisia. This teleconnection
between northwestern Europe and North Africa is a classical synoptic configuration, which is well
known nowadays in southern France during "Mistral" and "Tramontane" winds. Northwesterly
air masses are canalized under the influence of an anticyclone located over Spain and a low
pressure centre over the southern Alps and Gulf of Genova. The winds accelerate and become
drier down the Rhone Valley (Mistral) and by invading the area between Pyrefi6es and Massif
Central mountains (Tramontane) so that their influence (when they are exceptionally strong) is
sometimes felt in Corsica, and even in North Africa. Perhaps this meteorological situation
occurred very often during the summer of 1816. However, the coldness of this year was not
general: apparently Morocco and southern Spain were largely influenced by southerly winds since
the negative anomalies are lower than 1°C. Central Europe, with a more continental climate, was
also less affected by this general cooling.

304

0° 10° E

Figure 7: The distribution of the annual temperature anomalies in Europe and northwestern Africa in
1816.

Conclusions

The summer of 1816 was the coldest in the last three hundred years in central Canada. The other
seasons have been about normal or slightly colder. The greatest negative anomalies concern the
area southeast of Hudson Bay, and the smallest ones the region north of Hudson Bay. In Europe,
1816 was among the coldest years of the millennium with minimum temperatures (anomalies
close to -3°C) extending from the British Isles to Tunisia. Northwesterly winds chilled western
Europe: these cold air masses accelerated and dried between the Pyr6n6es and Massif Central
mountains (Tramontane wind) and down the Rhone Valley (Mistral wind), crossing the
Mediterranean Sea to Tunisia. At the same time, Morocco remained relatively warm. In central
Canada, more details are available about 1816 from seasonal records. Apparently the cooling
began in winter in the Southwest and ended by the close of summer, whereas it began a season
later (in spring) in the East, also ending later (in autumn).

The severe climate characteristic of this year must be placed in context. The cooling began a few
years before 1816, at the beginning of the decade. Then aerosols from the volcanic eruption of
Tambora only exacerbated a trend already existing in Europe and central Canada. Comparison
of results from the two continents shows strong coherency in the low-frequency variations of
temperature on both sides of the Atlantic Ocean during this globally-cold period of the Little Ice
Age. The coherency is weak in warmer periods.

305

This study shows how to synthesize various proxy series available to provide a better knowledge
of past climatic changes in Europe. More records must be used in order to obtain maximum
reliability of the gridded temperature reconstructions. It also appears that the Mediterranean
climate, which is now very different from the northern European one, has been so for many
centuries. Information concerning northern Europe cannot be directly extended to southern
Europe. More proxy series related to the Mediterranean climate must be collected to achieve
better reliability.

References

Alexandre, P. 1987. Le climat en Europe au Moyen-Age. Ecole des Hautes Etudes en Sciences
Sociales. Paris. 825 pp.

Allen, W.T.R. 1977. Freeze-up, break-up and ice thickness in Canada. Fisheries and
Environment Canada, Atmospheric Environment Service CLI-1-77: 1-185.

Baillie, M.G.L. 1977. The Belfast oak chronology to 1001. Tree-Ring Bulletin 37:1-12.

Ball, T.F. and R.A. Kingsley. 1984. Instrumental temperature records at two sites in central
Canada: 1768 to 1910. Climatic Change 6:39-56.

Bergthorsson, P. 1969. An estimate of drift ice and temperature in 1000 years. Jdkull 19:94-101.

Bray, J.R. 1982. Alpine glacial advance in relation to proxy summer temperature index based
mainly on wine harvest dates, 1453-1973. Boreas 11:1-10.

Briffa, K.R., P.D. Jones and F.H. Schweingruber. 1988. Summer temperature patterns over
Europe: a reconstruction from 1750 based on maximum latewood density indices of
conifers. Quaternary Research 30:36-52.

Catchpole, A.J.W. and D.W. Moodie. 1975. Changes in the Canadian definitions of break-up
and freeze-up. Atmosphere 12:133-138.

Cropper, J. P. and H.C. Fritts. 1981. Tree-ring width chronologies from the North American
Arctic. Arctic and Alpine Research 13:245-260.

Dansgaard, W., S.J. Johnson, H.B. Clausen and C.C. Langway Jr. 1971. Climatic record
revealed by the Camp Century ice cores. In: The Late Cenozoic Glacial Ages.
K.K. Turekian (ed.). Yale University Press, New Haven, pp. 37-56.

Efron, B. 1979. Bootstrap methods: another look at the jackknife. The Annals of Statistics 7: 1-26.

Fritts, H.C. 1976. Tree-Rings and Climate. Academic Press, New York. 567 pp.

Fritts, H.C. and J. Guiot. 1988. Methods for calibration, verification and reconstruction. In:
Methods in Tree-Ring Analysis: Applications in the Environmental Sciences. L. Kairiukstis
and E. Cook (eds.). Reidel, Dordrecht, (in press).

306

Guiot, J. 1985a. Reconstruction of seasonal temperature and sea-level pressure in the Hudson Bay
area back to 1700. Climatological Bulletin 19:11-59.

. 1985b. The extrapolation of recent climatological series with spectral canonical

regression. Journal of Climatology 5:325-335.

. 1986. Reconstructions des champs thermiques et barom&riques de la region de la Baie

d'Hudson depuis 1700. Environment Canada, Atmospheric Environment Service,
Canadian Climate Centre Report 86-1 IF (francais) Rapport 86-1 IE (English): 111 pp.

Hansell, R.I.C. (Editor). 1984. Study on temporal development of subarctic ecosystems -
determination of the relationship between tree-ring increments and climate. Report to
Department of Supply and Services Canada, Atmospheric Environment Service,
Downsview. Contract OSU84-00041.

Hollstein, E. 1965. Jahrringchronologische von Eichenholzern ohne Waldkande. Bonner
Jahrbilcher 165:12-27.

Huber, B. and V. Giertz-Siebenlist. 1969. Unsere tausendjahrige Eichen-Jahringchronologie
durchschnittlich 57 (10-150) fach belegt. Sitzungsberichte der Osterreichischen Akadamie
der Wissenschaften 178:37-42.

Ingram, M.J., D.J. Underhill and T.M. Wigley. 1978. Historical climatology. Nature
276:329-334.

Jones, P.D., S.C.B. Raper, B. Santer, B.S.G. Cherry, C. Goodess, P.M. Kelly, T.M.L. Wigley,
R.S. Bradley and H.F. Diaz. 1985. A grid point surface air temperature data set for the
northern hemisphere. United States Department of Energy, Contract
DE-AC02-79EV 10098, Washington, TR022. 256 pp.

Lamb, H.H. 1982. Climate, History and the Modern World. Methuen, London. 387 pp.

Legrand, J. P. 1979. Les variations climatiques en Europe Occidentale depuis le Moyen-Age
(suivi d'une note historique de LeRoy Ladurie). La Mitiorologie, Vfime serie I 61 :67-182.

LeRoy Ladurie, E. and M. Baulant. 1981. Grape harvests from the 15th through the 19th
century. In: Studies in Interdisciplinary History. R.I. Rotberg and T.K. Rabb (eds.).
Princeton University Press, New Haven, pp. 259-269.

Parker, M.L., L.A. Jozsa, S.G. Johnson and P. A. Bramhall. 1981. White spruce annual ring-
width and density chronologies from near Great Whale River (Cri Lake), Quebec. In:
Climatic Change in Canada 2. C.R. Harington, (ed.). Syllogeus 33:154-188.

Pfister, C. 1981. An analysis of the Little Ice Age climate in Switzerland and its consequences
for agriculture production. In: Climate and History. T.M.L. Wigley, M.J. Ingram and
G. Farmer (eds.). Cambridge University Press, Cambridge, pp. 214-248.

Pilcher, J.R. and M.G.L. Baillie. 1980. Eight modern oak chronologies from England and
Scotland. Tree-Ring Bulletin 40:45-58.

307

Rannie, W.F. 1983. Break-up and freeze-up of the Red River at Winnipeg, Manitoba, Canada
in the 19th century and some climatic implications. Climatic Change 5:283-296.

Richter, K. 1988. Dendrochronologische und Dendroklimatologische untersuchungen an Kiefern
(Pinus sp.) in Spanien. Ph.D. thesis, Universitat Hamburg, Hamburg.

Serre, F. 1978. The dendoclimatological value of the European larch (Larix decidua Mill.) in the
French Maritime Alps. Tree-Ring Bulletin 38:25-33.

Serre-Bachet, F. 1985. Une chronologie pluris^culaire du sud de l'ltalie. Dendrochronologia
3:45-66.

. 1986. Une chronologie maitresse de sapin {Abies alba Mill.) du Mont Ventoux (France).

Dendrochronologia 4:87-96.

Tessier, L. 1981. Contribution dendroclimatique a la croissance 6cologique du peuplement
forestier des environs des chalets de l'Orgere. Travaux Scientifiques du Pare de la Vanoise
12:9-61.

Till, C. 1985. Recherches dendrochronologiques sur le cedre de l'Atlas (Cedrus atlantica (Endl.)
Carriere) au Maroc. Ph.D. thesis, UCL, Louvain-la-Neuve. 231 pp.

Till, C. and J. Guiot. 1988. Reconstruction of precipitation in Morocco since 1100 based on
Cedrus atlantica tree-ring widths. Quaternary Research, (in press).

Williams, L.D. and T.M.L. Wigley. 1983. A comparison of evidence for late Holocene summer
temperature variations in the northern hemisphere. Quaternary Research 20:286-307.

308

Climatic Conditions for the Period Surrounding the Tambora Signal in
Ice Cores from the Canadian High Arctic Islands

Bea Taylor Alt1, David A. Fisher2, and Roy M. Koerner1

Abstract

The Tambora volcanic signal (acid layer) has been identified in ice cores taken from Agassiz and
Devon ice caps in the Canadian High Arctic. Oxygen isotope values (representative of annual
precipitation temperature) and melt percent values (representative of summer temperatures) from
core sections surrounding the volcanic signal have been examined in detail and compared with
present day conditions. The results suggest that the Tambora volcanic eruption did not produce
significant cooling in the Canadian High Arctic.

On Agassiz Ice Cap the ice representing the year after the volcanic signal shows an increase
(warming) of both oxygen isotope and melt percent values followed by a return to pre- volcanic
conditions. On Devon Ice Cap the oxygen isotope values began to decrease (cool) prior to the
Tambora signal and cool to a minimum 25 years later. Melt percent values on Devon Ice Cap had
already reached a minimum by the time of the eruption and this persisted for 45 years.

Based on modern synoptic studies, the circulation pattern during the summer season containing
the Tambora signal (1816) is best represented by the 1972 analogue. In this analogue a long,
narrow vortex at 500mb (50kPa) extends from the Siberian side of the central Arctic Ocean
across the Pole deep into Labrador-Ungava, and is held tight against Greenland by a strong ridge
of high pressure in the Alaska-Beaufort Sea area. This pattern results in strong cold northwesterly
flow, with frequent light precipitation and very little melt on the ice caps. The pattern is broken
occasionally by the joining of the Alaska and Greenland ridge which brings clear skies and some
melt to the islands along the northwestern edge of the archipelago.

Introduction

The records from deep ice cores extracted from ice caps in the Canadian Arctic Islands provide
insight into the climatic conditions in the islands during the decade surrounding the eruption of
Mount Tambora in Indonesia. It is also possible from these data to address the question of
whether single volcanic events (such as the Tambora eruption) produce significant deviations in
proxy annual temperatures and/or proxy summer temperatures in this area of the High Arctic.
Using modern synoptic-climate analogues, inferences can be made about synoptic circulation
conditions at the time of the eruption of Mount Tambora.

Terrain Sciences Division, Geological Survey of Canada, 601 Booth Street, Ottawa, Ontario K1A 0E4, Canada.
Department of Glaciology, Geophysical Institute, University of Copenhagen, Haroldsgade 6, DK-2200,
Copenhagen N, Denmark.

Geological Survey of Canada Contribution No. 156789.

309

For this study the most complete data are available from a core drilled in 1984 at the top of a
local dome on the Agassiz Ice Cap (A84) on northern Ellesmere Island (Figure 1). Results from
another Agassiz Ice Cap core, drilled in 1977, 1.2 km down the flow line from the dome (A77)
and from a combined time series of three cores taken from the top of Devon Ice Cap (Figure 1)
in 1971, 72 and 73 (referred to as D123) are also examined.

Figure 1: Location of ice cap deep core drill sites in the Canadian Arctic Islands and Greenland.

These cores have been extensively discussed elsewhere (Paterson et al. 1977; Koerner 1977;
Koerner and Fisher 1981; Fisher et al. 1983; Koerner and Fisher 1985; Fisher et al. 1985; Alt
1985; Alt et al. 1985; Fisher and Koerner 1988). Here we will confine ourselves to the
parameters and analyses which come closest to providing climatic data with an annual resolution.
Every attempt has been made to provide an accurate annual time scale and to make these
consistent for the period surrounding the Tambora eruption. It should be noted that northern
hemisphere eruptions deposit acidic aerosols on the snow within months of the event, but
southern hemisphere acid signatures first appear as much as one year after the southern eruptions.

310

Melt Features and Oxygen Isotopes

Melt layers (ice formed by the refreezing of meltwater) can be identified in ice cores by their
relatively low concentration of air bubbles. In the upper reaches of an ice cap this melt is
indicative of summer warmth (Koerner 1977). It is expressed as either melt-layer thickness (m)
or as a percent of the total annual accumulation (PC). As melt (m or PC) can never be less than
0, possibly very severe summers are not adequately represented in the melt record. Table 1 gives
the size of errors associated with time series of PC.

5(lsO) is the ,60/l80 ratio expressed as the fractional difference between the ratio in the sample
and the ratio in "standard mean ocean water" (SMOW) measured in percent. In polar snow, b
is negative. Initially, b was used as an indicator of mean annual temperature due to its dependence
on the temperature at which condensation takes place. However up to seven non-temperature
effects can alter the 6 at a given site (Dansgaard et al. 1973; Fisher 1979; Fisher and Alt 1985;
Johnsen et al. 1989). For the present discussions the b values should be viewed as representing
the mean annual precipitation temperature (temperature during precipitation events). Table 1 gives
the size of the errors associated with time series of b.

The Tambora Volcanic Signal

The volcanic peaks are identified by measuring the electrical conductivity (ECM) of core
segments using brass electrodes with a 1250 DCV potential between them. The resulting values
are plotted on a time scale derived from models and measurements of annual layer thickness
(Koerner and Fisher 1985). Major acid layers are correlated with those in the absolutely dated
Dye 3 core (Hammer 1980). The Tambora volcanic signal appears in the Dye 3 core in 1816.

In the A84 core the Tambora signal was identified in core 16 (Figure 2, bottom left). This core
segment is in the firn at a depth of 27 m, is 146 cm long and represents approximately 16 years
of accumulation. The core segment containing the peak signal from the eruption of Mount Laki
in Iceland which occurs at 30 m depth in the ice is shown in Figure 2, bottom right. The peak
ECM value for this Icelandic volcano is much higher than that from the Indonesian volcano,
Tambora.

The melt-layer thickness values m for these cores have been plotted in a manner consistent with
the ECM values. The melt and acid feature data are lined up within 5 cm. The Tambora event
falls between two melt features whereas the Laki acid layer comes 40 cm above the big melt
feature in core 18 (Figure 2, top).

Time Series of Acid, Melt and Oxygen Isotope Values for Agassiz

The annual average values of ECM, PC and b have been plotted on the volcanic time scale for
the Agassiz 84 core (Figure 3). This is probably the best time scale of the various Agassiz Ice
Cap cores, and great care has been taken to align the three stratigraphies, however discrepancies
of a year are possible. The Laki signal is absolutely dated as 1783 and the Tambora signal as
1816.

Both the b and PC values (Figure 3) reach a maximum just following Tambora. The average
annual b has been calculated for the period of 1941-70, which is used as a meteorological normal.
Compared to this, the 1816 b value is slightly above the modern normals. The 1941-71 melt
normals could not be calculated for A84. Instead the mean PC (dotted line, Figure 3) and the

311

6.6

1.2 -
1.0-
.8
.6-
.4-

.2-
0

.4 _

.3 -

.2 -

.75

CORE 15 4*- CORE 16— +\

Figure 2: Melt-layer thickness (m) shown as actual observations of the thickness of individual layers of
ice (top) and volcanic electrical conductivity measurements ECM values from the Agassiz 84
core segments containing the Tambora and Laki volcanic signals (bottom). Values are plotted
on a depth scale with top of the core to the left. More than one melt layer can occur in a year.
Core 16 is 146 cm long, so 5 cm is equivalent to the smallest horizontal increment on the ECM
plot.

312

Table 1: Errors in Percent Melt PC and Oxygen Isotope b Time Series.

Site

Interval

Start AD

SD PC

SD b

Accumulation

years

average

noise
1 yr 5 yr

% %

average

noise
1 yr 5 yr

% %

(ice)
cm/yr

All

500

1946

2.8

>10 2.5

-31.5

0.48 0.35

17.5

A84

800

1961

4.1

>25 5.5

-28.5

0.32 0.23

9.8

Devon1

500

1956

7.0

> 8 1.6

-28.0

0.55 0.40

23.0

1 Devon combined record; 6(73+72) and PC(71 +72 + 73).

Note: SD is the standard deviation. The Devon SD(noise) data has been measured, but the All and A84 noise data
is estimated.

Eureka mean July temperature (5.7 °C) for the 195 1-60 decade have been calculated. The 1941-70
Eureka July normal temperature (5.4°C) is slightly colder than the 1951-60 decade. The 1816
melt is also below the 1951-60 decade mean or probably near the modern normal, whereas the
1817 melt is considerably greater - comparable to the 1951-60 warm period.

When the A84 values are plotted as five-year averages (Figure 4) the b profile might well be
interpreted as showing a cooling immediately following Tambora. This is in direct contrast to the
annual averages that show an immediate warming. Care must be used, therefore, in interpretation
of averaged values in studying the short-term effects of single volcanic eruptions.

As mentioned, the A84 core has the most accurate time scale but it is at the top of a dome. Here
the light winter snow is consistently scoured (i.e., blown away). This results in a 6 record which
is "warmer" than it would be if the winter snow was included. The A77 core lies sufficiently
downslope from the dome to escape the scouring effect. A detailed plot of annual average 5
values for the Tambora period from A84 and A77 (Figure 5) shows that the minimum preceding
the Tambora signal is much colder in the unscoured core than in the scoured (A84) core. Based
on the most recent time scale for the A77 core, this puts 1816 near the bottom of this minimum
followed by a rise to 1970 normal values by 1819.

Time Series from Other High Arctic Cores

From the Devon blended 1971, 72 and 73 core records (D123) only five-year averages of 5 and
melt percent are available (Figure 6). The time scale is the vertical velocity time scale fine-tuned
by analysis of annual layering as deduced by seasonal swings in microparticle concentrations and
radiocarbon dating of gas bubbles in the ice (Paterson et al. 1977) and corrected for the location
of the well-marked Laki eruption in the meltwater electrolytic conductivity records (Koerner and
Fisher 1981). It is accurate to within a few years at the 1816 level. The 1941-70 normals are
shown for both 5 and melt percent (Figure 6).

313

Figure 3: Annual averages from the Agassiz 84 core (A84): (a) oxygen isotope values 5; (b) melt
expressed as percent of accumulation PC; and c) volcanic ECM values. The arrows on the
ECM plot show the magnitude of the actual measured peak ECM value for various volcanic
signals. The dotted line shows the 1951-60 PC decade average. The dashed line shows the
1941-70 8 30-year average.

314

Figure 4: Five-year averages from A84 of oxygen isotope b (dashed line 1941-70 average) and melt
percent PC.

The most striking feature in the Devon Ice Cap blended record is the very cold summer period
indicated by consistently low melt values during the whole period 1810-55. The eruption of
Tambora occurred well after the beginning of this period. The Laki eruption, on the other hand,
occurred during a period of increasing melt which reached almost to present normal values by
1800.

On the 5 plot 1816 falls on a cooling trend which began before 1810 and reaches its lowest values
for the 200-year period during the 1830s.

Comparing the 10-year melt averages for A77, D123 and Dye 2 on Greenland (Figure 7), we see
that 1816 falls in a period of generally cold summers at all sites. In all cases the cold period
began before the eruption of Tambora. At Dye 2 the lowest melt values occur in the 1820s and
30s but the long flat cold period of the D 123 cores is not present.

The Effect of Single Volcanic Events on High Arctic Climate

None of the ice core records examined above shows definitive evidence of cooling resulting from
the eruption of Mount Tambora. Those records which reach a minimum at some time following
1816 all show a cooling trend beginning before the Tambora volcanic signal. The same is true
of the Laki eruption. Two other volcanic signals have been identified in the A84 core (Figures
3 and 4), dated by correlation with the Dye 3 volcanic record and also plotted on the D123

315

r-26

Figure 5: Detailed comparison of the annual mean A84 and A77 oxygen isotope values 6 and the A84
melt percent values PC from around the Tambora volcanic signal. The five-year mean values
(solid lines) and the 30-year 6 normals from 1941-70 (dashed lines) are shown for the oxygen
isotope values. The dotted line shows the average PC values for 1951-60.

316

record (Figure 6). The year of the eruption of Mount Agung in Indonesia is also shown on these
figures. The Agung signal has not been positively identified in the Canadian cores as it was not
sufficiently acidic. The five-year A84 PC averages (Figure 4) suggest cooling following Katmai,
but close examination of the annual melt record (Figure 3) shows the season prior to Katmai was
also cold. Both melt and b values show high values following the Krakatau signal. On Devon Ice
Cap the five-year averages for Krakatau drop sharply in summer melt but the b values are already
low. The Agung eruption appears to occur at the bottom of a 5 minimum in the A84 core. Melt
data are not available past 1961. In the Devon cores the eruption follows a b minimum and is on
a well-established downward melt trend.

Figure 6: Five-year averages for D123, the Devon blended record (1971, 72 and 73 cores), of oxygen
isotope b and melt percent PC. The 30-year normals, 1941-70 are indicated (dashed lines).

The very cold summer of 1964 in the Canadian High Arctic, and the subsequent generally lower
summer temperatures have been attributed to the effects of dust from Agung (Bradley and
England 1978). Close examination of the hemispheric temperature plots of Dronia (1974) and
Kelly et al. (1982), Figure 8, show that in both cases the hemispheric temperatures had begun
to cool long before the eruption of Agung in 1963. The rather dramatic drop of July mean
temperatures seen in the plots from the northern Canadian Arctic Island stations (Figure 9) is,
in fact, a result of the record high temperatures in the 1962 season.

317

These results do not appear to indicate that single volcanic eruptions cause lower summer or
annual temperatures in the northern Canadian Arctic Islands. This does not rule out the possibility
that multiple eruptions in a period could have a cumulative effect on temperatures (Hammer et
al. 1980) or that single volcanic events produce abrupt, short-lived temperature depressions on
a hemispheric scale (Bradley 1988). Single events could also be responsible for significant
anomalies in the atmospheric circulation regime in the Canadian Arctic Islands such as occurred
in the summer of 1964 (Alt 1987).

Summary of Core Results

Now we can review what the ice core analyses reveal about climate in the area at the time of the
Tambora volcanic signal. The results are expressed in Table 2 as simple estimates of the
temperature anomalies with respect to the modern normals (1941-70).

On the Agassiz Ice Cap the summer melt conditions, and thus the summer temperatures, were
near or slightly below the 1941-70 normals. There was a rise in the melt values immediately after
the Tambora event to values similar to the relatively warm 1951-60 decade as experienced at
Eureka.

The annual temperature (or more accurately the annual precipitation temperature) on Agassiz Ice
Cap appears to have been lower than the 1941-70 normals. The scoured A84 core shows the
Tambora signal to be part of a slight rise from below-normal conditions to above 1941-70
normals. The unscoured core A77 shows 1816 to be on a warming trend from a very cold period.

On Devon Ice Cap there was very little summer melt, indicating very cold conditions. These
conditions began around 1810 and persisted until the late 1850s. This is the longest very cold
period in the 800-year record.

On Devon Ice Cap the Tambora signal falls on a cooling trend of annual (or precipitation)
temperature beginning about 1810, when the oxygen isotope values were very near the modern
normals. This cooling trend could be viewed as part of a general decline beginning before the
time of the Laki signal.

Table 2: Conditions on Canadian Arctic Island Ice Caps During the Period of the Tambora Volcanic
Signal as Deduced From Ice-Core Records.

Season Ice Cap Temperature Remarks

SUMMER Agassiz 0 (normal) then rising

(from melt
percent records)

Devon

- - (very cold)

already very cold

ANNUAL

Agassiz

- (cold)

on a warming trend

(from oxygen
isotope values)

Devon

- (cold)

on a cooling trend

318

PC 10yr

15-

-TJ

la-

Dye 2

i i i l i i i i I i i i i l i i i i I i i i i I

9 58 100 158 280 258

/ears

Figure 7: Comparison of 10-year averages of A77, Devon blended D123 and Dye 2 (Greenland) melt
percent values.

319

Synoptic Conditions

Based on the core results for the period around the Tambora signal it is now possible to examine
the synoptic circulation patterns which would be expected to produce these conditions on the two
ice caps. Previous studies of synoptic analogues and ice-core results (Alt 1985; Alt et al. 1985;
Alt 1987) have suggested that this period of the Little Ice Age was dominated by summers similar
to the summer of 1972 (Figure 10). The most important feature of the 1972 circulation analogue
for the study area is the persistence of a long deep 500mb (50kPa) vortex held against Greenland
by a strong ridge over Alaska and western Canada. This produces persistent northwesterly flow
into the Canadian Arctic Islands from the central Arctic Ocean and a deep layer of very cold air
in the northern Baffin Bay area.

°c

- 1.0

- 0.0
--1.0
--2.0

1880

1900 1920 1940 1960

— r
1980

-3.0

— i —

1950

— I —
1960

1970
j

°C

- 0.6

- 0.4

- 0.2
0.0

—0.2
--0.4
--0.6

Figure 8: Two depictions of the annual temperature record for the arctic: (a) annual temperature
departures from the 1946-60 reference period for 65-85°N (after Kelly et al. 1982); and
(b) annual deviations from the 25-year mean 1949-73 of thickness of the 500/1, OOOmb layer for
65-90°N (after Dronia 1974).

320

-2 '- + + + + A + + + + + + + + + +

1950 52 54 56 58 6 0 62 64 66 68 70 72 74 76

Figure 9: Normalized deviation from the mean of July temperature for Canadian Arctic Islands stations
[(July mean - July normal)/July standard deviation] from P. Schofield (personal
communication).

These features are evident from comparison of the mean July 500mb (50kPa) height contours for
the period 1948-78 with those of 1972 (Figure 1 la,b). We see that the 1972 vortex is deeper and
narrower than the mean, and shifted eastward from the mean position by a ridge over the
Beaufort Sea. The flow into the High Arctic Islands is stronger than normal as seen by the closer
spacing of the contour lines. There is also a strong ridge over the Barents Sea. The mid-latitude
circulation during the entire 1971-72 season was stronger than usual and distinctly meridional
(i.e., with strong north-south components).

321

Figure 10: Schematics of summer synoptic characteristics for the Agassiz and Devon core-site area for
various periods of the last 800 years. These were deduced by applying modern analogues to
the ice-core results; represented here by the five-year averages of oxygen isotope b and melt
percent PC from Devon Ice Cap (Alt 1985).

322

Figure 11: Mean July pattern of 500mb (50kPa) height contours in decameters (dam). In order to focus
on the polar vortex, contours up to 560 dam only are shown: (a) for 1948-78 (after Harley
1980); and (b) for 1972.

323

In order to examine the surface weather conditions associated with the 1972 anomaly, the actual
synoptic charts for 3 July 1972 are shown in Figure 12a, b. Here we see the long upper vortex
(Figure 12a) with multiple centres, one over Labrador-Ungava, a second in northern Baffin Bay
and a third over the Pole. This supports a surface trough (Figure 12b) across northern Ellesmere
Island down the west coast of Greenland to Davis Strait. The strong northwesterly flow, indicated
by the closely packed isolines, extends across the Canadian Arctic Islands into Keewatin, Hudson
Bay and James Bay at all levels from the surface to 500mb (about 5,000m). Cold moist air from
the central Arctic Ocean is pushed south into these areas. In the northern islands extensive low
cloud, high humidity and frequent light precipitation accompany these conditions. Over Devon
Ice Cap precipitation may be enhanced by the persistence of the northern Baffin Bay low, which
picks up additional moisture from the open water.

Both the mean July 1972 (Figure 1 lb) and 3 July 1972 (Figure 12a and b) patterns appear to be
consistent with conditions proposed for the Hudson Bay region during 1816. Wilson's (1983)
studies show prevailing NW-NE winds in June and July 1816 and also suggest as a modern
analogue the summer of 1972. High pressure west of Hudson Bay (in the case shown here, an
extension of the Alaska ridge) is an important feature of her proposed 1816 circulation patterns.

Figure 12 (a): Synoptic chart for 3 July 1972: 500 mb (50 kPa) height contours in decametres (dam).

324

Figure 12 (b): Synoptic chart for 3 July 1972: surface pressures in mb.

Figure 12a,b also gives us an indication of the synoptic conditions that previous detailed synoptic
studies of summers from 1960-78 have shown could produce melt on Agassiz Ice Cap and not
on Devon Ice Cap. Close inspection shows that the ridge extending from the Beaufort Sea to
southern Manitoba (which is responsible for keeping the trough or vortex tight against Greenland)
can be seen pushing northeast across the northern Canadian Arctic Islands toward the north
Greenland ridge. If these join (as they did later in July 1972), the vortex is cut off, becoming a
closed low. The closed system can lie over Labrador-Ungava or farther west, as happened in July
1972. Along the northwestern edge of the islands the ridge results in subsidence through the
whole troposphere, which dissipates the cloud and fog. Melt is produced by increased solar
radiation (sometimes aided by warm-air advection) over the northern and western ice caps. Devon
Ice Cap, however, is often under the influence of the cyclonic circulation in Baffin Bay and not
as likely to experience melt. In fact, on Devon Island summer accumulation (snow) may occur
under these conditions. These ridging conditions are often brief but they can produce significant
melt on the ice caps and a touch of summer in the northwestern islands, as happened in mid-July
1972 (Alt 1987).

We can also say that this pattern resembles the mean winter conditions, and suggest that in years
of this kind the winter circulation is never really broken down. The temperature gradients remain
strong as do the mid-latitude westerlies. Summer comes only briefly to the ice caps if and when
the blocking ridges join across the northern islands. These ridging conditions are more effective
in the northwestern islands and may not produce any melt at the core site on Devon Ice Cap.

325

Conclusions

The Tambora signal can be identified as an acid layer in the Agassiz 84 core. The oxygen isotope
5 and melt values PC, even allowing for a one-year discrepancy and other considerations such
as noise and scouring, do not show evidence of cooling due to the eruption of Mount Tambora,
although it may have occurred part way down a cooling trend. Nor is there definitive evidence
of cooling in the northern Canadian Arctic Islands following the eruption of Laki, Krakatau,
Katmai or Agung.

On Agassiz Ice Cap, conditions in the year dated as 1816 were near, or somewhat below, modern
normals (1941-70) but rise to a secondary peak immediately following the Tambora signal. Care
must be taken when interpreting average values for periods longer than a year as they can easily
obscure the short-term variations. However, on the Devon Ice Cap blended five- year average
plots, 1816 falls in a prolonged period of very low summer melt and below modern normal 5
levels (annual or precipitation temperature); both of which began about 1800. Similarly the Dye 2
10-year average plot shows Tambora occurring on a well established cooling trend.

The climatic conditions suggested by the ice-core analyses around the Tambora eruption strongly
resemble those of the summer of 1972, which has been identified as the modern analogue for
melt suppression on High Arctic ice caps. This pattern, which features a long deep upper vortex
extending from the Siberian side of the central Arctic Ocean across the eastern Canadian Arctic
Islands to Labrador-Ungava and a strong ridge from the Beaufort Sea and Alaska into the
prairies, appears to be compatible with synoptic interpretations from other parts of Canada. This
pattern represents an intensification of the conditions which appear to have dominated the latter
part of the Little Ice Age in the High Arctic islands.

References

Alt, B.T. 1985. A period of summer accumulation in the Queen Elizabeth Islands. In: Critical
Periods in the Quaternary Climatic History of Northern North America. Climatic Change
in Canada 5. C.R. Harington (ed.). Syllogeus 55:461-479.

. 1987. Developing synoptic analogues for extreme mass balance conditions on Queen

Elizabeth Island ice caps. Journal of Climate and Applied Meteorology 26(12): 1605-1623.

Alt, B.T., R.M. Koerner, D.A. Fisher and J.C. Bourgeois. 1985. Arctic climate during the
Franklin Era as deduced from ice cores. In: The Franklin Era in Canadian Arctic History.
Pat Sutherland (ed.). National Museum of Man, Mercury Series, Archaeological Survey of
Canada Paper 131:69-92.

Bradley, R.S. 1988. The explosive volcanic eruption signal in northern hemisphere continental
temperature records. Climatic Change 12:221-243.

Bradley, R.S. and J. England. 1978. Volcanic dust influence on glacier mass balance at high
latitudes. Nature 271:736-738.

Dansgaard, W., S.J. Johnsen, H.B. Clausen and N. Gundestrup. 1973. Stable isotope glaciology.
Meddelelser om Grenland 197(2): 1-53.

326

Dronia, H. 1974. Uber Temperaturanderungen die frier Atmosphare auf der Nordhalbkugel in
den letzten 25 Jharen. Meteorologische Rundschau 27:166-174.

Fisher, D.A. 1979. Comparison of 105 years of oxygen isotope and insoluble impurity profiles
from Devon Island and Camp Century ice cores. Quaternary Research 1 1(3):299-305.

Fisher, D.A., R.M. Koerner, W.S.B. Paterson, W. Dansgaard, N. Gundestrup and N. Reeh.
1983. Effect of wind scouring on climatic records from ice-core oxygen-isotope profiles.
Nature 301(5897):205-209.

Fisher, D.A., R.M. Koerner, N. Reeh and H.B. Clausen. 1985. Stratigraphic noise in time series
derived from ice cores. Annals of Glaciology 7:76-83.

Fisher, D.A. and B.T. Alt. 1985. A global oxygen isotope model - semi empirical, zonally
averaged. Annals of Glaciology 7:117-124.

Fisher, D.A. and R.M. Koerner. 1988. The effects of wind on 6(018) and accumulation give an
inferred record of seasonal 6 amplitude from the Agassiz Ice Cap, Ellesmere Island,
Canada. Annals of Glaciology 10:34-38.

Johnsen, S.J., W. Dansgaard and J. W.C. White. 1989. The origin of Arctic precipitation under
present and glacial conditions. Tellus Series B, 41B(4):452-468.

Hammer, C.U. 1980. Acidity of polar ice cores in relation to absolute dating, past volcanism,
radio-echoes. Journal of Glaciology 25(93):359-372.

Hammer, C.U., H.B. Clausen and W. Dansgaard. 1980. Greenland Ice Sheet evidence of
post-glacial volcanism and its climatic impact. Nature 288(5788):230-235.

Harley, W.S. 1980. Northern hemisphere monthly mean 50 kPa and 100 kPa height charts.
Environment Canada, Atmospheric Environment Service CLI-80:29.

Kelly, P.M., P.D. Jones, C.B. Sear, B.G. Cherry and R.K. Tavakol. 1982. Variations in surface
air temperatures. Part II: Arctic regions, 1881-1980. Monthly Weather Review
110(2):71-83.

Koerner, R.M. 1977. Devon Island Ice Cap: core stratigraphy and paleoclimate. Science
196(4285): 15-18.

Koerner, R.M. and D.A. Fisher. 1981. Studying climatic change from Canadian High Arctic ice
cores. In: Climatic Change in Canada 2. C.R. Harington (ed.). Syllogeus 33:195-215.

. 1985. The Devon Island ice core and the glacial record. In: Quaternary environments;

eastern Canadian Arctic, Baffin Bay and western Greenland. J.T. Andrews (ed.). Allen
and Uwin, Boston, pp. 309-327.

Paterson, W.S.B. and seven others. 1977. An oxygen-isotope climate record from Devon Island
Ice Cap, Arctic Canada. Nature 266(5602):508-51 1.

Wilson, C.V. 1983. The Little Ice Age on eastern Hudson Bay: summers at Great Whale, Fort
George, Eastmain, 1814-1821. Canadian Climate Centre Report No. 83-9:1-145.

327

Europe (including Iceland)

1816 - a Year without a Summer in Iceland?

A.E.J. Ogilvie1

Abstract

There has been considerable speculation as to whether the eruption of Mount Tambora in April
1815 caused a world-wide lowering of temperatures and a "year without a summer" in the
following year of 1816. In this paper, the weather during 1816 is detailed for one specific
location: Iceland. The weather data used are taken from documentary accounts written at 10
different sites in Iceland. These suggest that the winter and spring of 1816 were very cold and
unfavourable in most parts. The summer was mainly cold in the north, wet in the east and highly
variable elsewhere. Many accounts of the autumn focus on the variability of the weather.
Although it would seem that, on the whole, the summer weather was not sufficiently extreme for
this year to be termed a "year without a summer," adverse weather did cause some impact on
society. It seems very likely that there was direct climatic impact on important agricultural
practices such as the hay harvest and the growing of vegetables. The Arctic sea ice, although not
unusually heavy or prolonged in 1816, had a direct impact in northern Iceland, hindering fishing
and sealing. Indirect impacts on society are less easy to establish. However, it seems likely that
some social stress described in 1816 may be at least partly attributed to the climate.

Introduction

Although the precise nature of the effects of volcanic eruptions on the general circulation of the
atmosphere are, as yet, unknown, there can be little doubt that major volcanic eruptions do affect
the Earth's climate (e.g., Lamb 1970; Kelly and Sear 1984; Sear etal. 1987; Bradley 1988). The
possible effects of one very large eruption - that of Mount Tambora in April 1815 - has excited
particular interest. Although some researchers (e.g., Landsberg and Albert 1974) have concluded
that this eruption did not have significant climatic effects, others have provided convincing
evidence to show that the subsequent year, 1816, was anomalously cold in many places (Stothers
1984; Kelly et al. 1984). The year 1816 has even been termed the "year without a summer"
(Stommel and Stommel 1979).

In this paper, the weather during 1816 is considered for one specific location - Iceland. In order
to place the year in context, the general climate of Iceland is considered first, both for the
twentieth century, and in terms of climatic variations in the past. Possible climatic impact in
Iceland during 1816 will also be discussed.

The Present and Past Climate of Iceland

The Twentieth Century Context

Our knowledge of the climate of Iceland is derived from two main data sources. The principal
of these is modern instrumental data. By the late nineteenth century, around 20 observing stations
were in existence, and with the establishment of the Icelandic Meteorological Office (Vedurstofa

Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, U.K.

331

Islands1) in 1920, the number of stations grew. By 1955, there were 66. From 1966 onwards,
the number has varied between 120 to 130 (Einarsson 1976, pp. 12-13). Information from these,
and other observing stations in the North Atlantic and Polar regions, plus oceanographic data,
has enabled a general picture of key factors in the climate and weather of Iceland to be
established. These are summarized below. For more detailed discussions on this topic, see
Eyth(5rsson and Sigtryggson (1971) and Einarsson (1976).

Main Features of the Climate of Iceland

The principal features of Iceland's climate are determined by its location at the frontier zone of
two very different air masses; cold polar air from the north, and warmer maritime air from the
Atlantic. Depressions moving toward Iceland from the western Atlantic often slow down as they
near the southwestern corner of Iceland, thus maintaining a flow of mild Atlantic air over the
country. This process causes thaws in winter, and rain and cool temperatures in summer. When
these depressions cross Iceland and move toward Norway, a flow of polar air may take their
place and bring much colder weather, especially in the northern part.

The alternating cold and milder air masses that Iceland experiences at varying intervals, and for
different durations, are the prime cause of the variability of Iceland's climate. This variability is
exacerbated by the two major ocean currents which flow around the island; the cold East
Greenland polar current, and the warmer Irminger current. The Arctic drift ice also has
considerable influence on the climate of Iceland. Most noticeably, when the ice is present off the
coasts, both land and sea temperatures are lowered.

Although weather conditions in Iceland vary greatly, generally winters are mild compared with
other northern continental locations, and summers tend to be cool. Typical temperature ranges
during the winter months of December, January and February vary between -2 and 1°C. The
warmest summer month is generally July, with a mean temperature varying from around 8 to
11°C, depending upon location.

The Past Climate of Iceland: Introduction

Information about the climate of Iceland derived from modern data is augmented and amplified
by what is known of the past climate of Iceland. This is derived largely from documentary,
historical evidence: the nature and use of such evidence is discussed briefly below. Although we
cannot hope to gain as accurate a picture from documentary evidence of climate as from modern
instrumental data, such evidence can act as a guide to what may have occurred in the past when
no other data are available. To this end, proxy temperature variations based on the use of
historical documentary evidence have been derived by Bergthtfrsson (1969) and by Ogilvie
(1984a, 1986, 1990). Incidence of the sea ice off the coast of Iceland in the past has been
estimated by these same authors, and by Koch (1945).

In the sections below, probable variations in the past climate of Iceland from medieval times to
the early nineteenth century are outlined. Prior to this, the available data sources for this period
are discussed.

The Icelandic characters "b" and "d" (for "th") and all accents are retained wherever these are used in the original.

332

Data Sources

The accuracy of any proxy-temperature indicator will depend on the quality of the data used. To
ensure high quality of documentary evidence, all sources must be analyzed carefully in order to
establish their reliability. Key questions to ask here are whether the author was close in time and
space to the events described; if this is the case, then a source is much more likely to be reliable
than if he were not. For more detailed discussions on source analysis in general, see Bell and
Ogilvie (1978) and Ingram et al. (1978). For discussions of the analysis of Icelandic sources see
Vilmundarson (1972) and Ogilvie (1981, 1984a, 1990, 1991).

Iceland's climatic history may be traced back to early settlement times in Iceland (from about
A.D. 870 onwards). However, the quality and availability of climatic and weather data vary
considerably. For the period up to about 1 170, there are no contemporary documentary sources,
and only brief and sporadic comments on weather and climate may be found in existing sources.
For the thirteenth century, a few reliable sources give some indication of possible changes in
climate. Many more descriptions of weather and climate exist for the fourteenth century. The
fifteenth century and the first half of the sixteenth century are very poorly documented. Typical
sources for this period are certain sagas, the medieval annals and works of geographical
descriptions.

From the early seventeenth century onward, many more reliable documents become available.
For the early to mid-eighteenth century, there is extensive coverage from a variety of different
sources including annals, travel accounts, government reports and weather diaries. These give
information for most seasons in many different parts of Iceland. For the late eighteenth and early
nineteenth centuries, sources of climatic and weather information are very full and detailed.

The earliest quantitative observations taken in Iceland date from the mid-eighteenth century
(Eyb(5rsson 1956; Kington 1972). However, these, and subsequent late-eighteenth and nineteenth-
century observations only cover a few months or years. Continuous temperature observations
commence in 1846 (Sigfusd6ttir 1969). These were made at Stykkisholmur, in the west. For the
period 1820-54, observations of temperature were taken in Reykjavfk or the near vicinity by J6n
Porsteinsson (1794-1855). A part of this important series was subsequently lost for many years.
However, the missing data were recently found by Trausti Jdnsson of the Icelandic
Meteorological Office, and he is engaged in their analysis (J6nsson, personal communication).

The Climate of Iceland from Settlement Times to about 1600

Iceland was settled, primarily from Norway, in the late ninth and early tenth centuries.
Circumstantial evidence suggests a fairly mild climate around this time. A cold period may have
occurred from about 1180 to 1210, while from about 1211-32 the climate may have become
milder. An early geographical treatise written in approximately 1250 {The King's Mirror)
mentions much sea ice between Iceland and Greenland at this time, and refers to Iceland's cold
climate. However, it is difficult to draw firm conclusions from statements such as this. From
about 1280 to 1300 the climate seems to have been fairly cold. During the early years of the
fourteenth century, severe weather is mentioned only infrequently (in 1313, 1320, 1321 and
1323), so this period may have been mild. Milder weather may well have continued to past the
mid-fourteenth century. The years 1360-80 are likely to have been colder. Little information is
recorded from the 1380s. Only two severe years are noted for the 1390s. Evidently, 1412-70 was
mild, and the 1480s or 1490s were years of dearth, possibly caused by severe weather. However,

333

very little information is available for 1430-1560. Likely the latter part of the sixteenth century
was mainly severe. A detailed discussion of all medieval historical sources containing comments
on the weather and climate, together with an analysis of their evidence, may be found in Ogilvie
(1991).

The Climate of Iceland from 1601 to about 1850

The first and second decades of the seventeenth century were, overall, probably relatively mild.
The years 1620-40 were cold, but 1641-70 was distinctly mild. From 1671-90, temperatures were
colder. The 1690s were very cold. The early years of the eighteenth century were relatively mild,
especially the first decade. The 1730s, 1740s and 1750s were cold, especially the two latter
decades. The 1760s were somewhat milder, the 1770s cooler again. The period 1781 to 1820 was
cold on the whole. The year 1816 must be assessed in the context of this prevailing background,
with mainly cold conditions spanning most of the preceding four decades. From 1821 to 1841
the climate is likely to have been milder, while the 1840s were very mild. For a fuller account
of climatic variations in Iceland during the seventeenth and eighteenth centuries, see Ogilvie
(1981, 1984a, 1986, 1990).

The Weather in Iceland During 1816

Data Sources

In order to build up a clear picture of weather and climate during 1816, a number of sources
were selected for detailed analysis. The main sources used are letters written by "Sheriffs" or
government officials in the 20 or so different districts or Sysla (plural Sysslur) of Iceland. These
letters contain information on such topics as grass growth and hay crop, trade, health and disease.
They also report on the weather, sometimes in very great detail. One letter used here (from
Snaefellsnessysla, in the west) gives daily data, as well as seasonal summaries. The letters were
sent at least annually, sometimes more frequently, to the Danish government in Copenhagen.
Written in Danish, they are all unpublished, and are now kept in the National Archives in
Reykjavfk1.

For this discussion of 1816 weather, letters were chosen from nine different sites in Iceland. Use
was also made of one other source; an annal, called Brandsstadaandll . This was written by Bjorn
Bjarnason (1789-1859) at Brandsstadir, Blondudalur in Austur-Hunavatnssysla in the north. This
annal describes events that occurred in Iceland each year from 1783 to 1858, and includes
detailed weather descriptions. Other available sources were not used. It was felt, however, that
the sources used here were adequate to provide good regional coverage over the year. As
Iceland's climate is regionally quite variable (Eythtfrsson and Sigtryggsson 1971; Ogilvie 1984a)
it was essential to consider different parts of Iceland.

The sites at which these various sources were written are shown in Figure 1. The sources are
from (in the order given on the map): (1) Ketilsstadir in the district of Sudur-Mulasysla in the
east; (2) Gardur in Sudur-Pingeyjarsysla in the north; (3) Modruvellir in Eyjafjardarsysla in the
north; (4) Vidvflc in Skagafjardarsysla in the north; (5) Brandsstadir in Austur-Hunavatnssysla
in the north; (6) Grof in Snaefellsnessysla in the west; (7) Sfdumuli in Myrasysla in the west;
(8) Leira in Borgarfjardarsysla in the west; (9) Reykjavik in Gullbringusysla in the southwest;
(10) Vfk in Vestur-Skaftafellssysla in the southeast. Some use was also made of a letter written
at Grund, a site very close to Modruvellir.

All translations of sources are by the author.

334

1816 - The Evidence

Introduction

In Table 1 , a brief synopsis is given of comments on weather from the sources described above.
In the left-hand column, the place at which the letter was written is shown according to its
number on the map. The columns in the centre show the main characteristics of the seasons. The
term "winter" here refers to the period from mid-October of one year (1815) to mid-April of the
next (1816). "Spring" covers mid-April to mid-June, "summer" is mid-June to August, and
"autumn" is September to mid-October. The column on the right-hand side of Table 1 shows
descriptions of sea ice.

Winter

From Table 1, it may be seen that letters from most districts report a severe winter. Two of the
letters written in the north, at Modruvellir in Eyjafjardarsysla, and at Gardur in Adaldalur in
Audur-bingeyjarsysla, stated that the winter was, respectively "more than unusually severe" and
"very severe." Two other sources, one from Vfk in Vestur-Skaftafellssysla in the southeast, and
the other from Leir^ in Borgarfjardarsysla in the west, both noted that the winter was "very
severe." The writer of the account from Reykjavflc wrote that the winter was "severe with much
snow and frost."

Some sources stress the variability of the weather this winter. Thus, the letter from Vidvfk in
Skagafjardarsysla in the north reported that there was "much snow and alternating thaws and
sharp frosts," and the letter from Sfdumuli in Myrasysla in the west gives a similar account.
According to Brandsstadaanndll, the winter was also very severe, but there were some spells of
calm and good weather in between, for example, from 25 November to 15 December (1815) and
from 15 January to about 21 February. Interestingly, the letter from Ketilsstadir in Sudur-
Mulasysla in the eastern part of Iceland stated that the winter was merely average, although there
was much snow.

Spring

The spring of 1816 was also relatively cold in most districts. The letters written at Vidvfk and
Modruvellir in the north characterized the spring as "dry and cold" and "quite severe",
respectively. According to the latter, the severity took the form of persistent northerly winds,
frost and cold air. These the writer attributed to the presence of sea ice which lay off the northern
coasts all spring. The letter from Gardur does not contain a description of spring weather as
such, but does mention sea ice. This is stated to have been present from the beginning of March
to mid-June. The spring was said to be "unusually cold" in the letter from Reykjavik.

According to Brandsstadaanndll, April was severe, but the weather improved at the end of the
month. The account from Leira\ in the west, also notes severe cold in April, but says that from
the end of the month to mid-May the weather was mild. From then it became cold again with
northeasterly winds, sleet and night frost to about 24 June. The Sheriff of Skaftafellssysla, writing
at Vfk, recorded a severe April and a mild May. The first 10 days of June were dry and frosty.

Summer

The weather during the summer of 1816 in Iceland was quite variable regionally. The northern
sources used make it clear that, in most northern areas, the weather was very poor. Most
accounts from the south and west report a mixture of both favourable and unfavourable weather.
The eastern source used here states that the summer was wet. We may look at accounts from
these regions in more detail, starting with the north. A summary of the data may also be found
in Figure 2.

335

25° 24° 23° 22° 21° 20° 19c

18" 17v

16l

15v

13^

km 100
-i * ■ '

I ' 1 1

miles 60

J L

24^

23^

22^

21*

20^

18l

17v

16v

15v

14v

1. Ketilssta5ir , SuSur-Mulasysla

2. GarSur, Su5ur-f>ingey jarsysla

3. M65ruvellir, Ey jaf jarSarsysla

4. Vi5vik, Skagaf jarSarsysla

5. BrandsstaSir , Austur-Hunavatnsyssla

6. Grof, Snsf ellsnessysla

7. Si5umuli, Myrasysla

8. Leira, Borgarf jar5arsysla

9. Reykjavik, Gullbringusysla
10. Vik, Vestur-Skaf taf ellssysla

Figure 1: Sites of sources used for 1816 weather reconstruction.

336

z
s

b
<

O
Z

W
H
Z

V
60

I
>

-o
a

2 §

<u *

w «

u en

> -3

u a u

u > «
o

o
2

o
y

■3
Q

a,—

I73
y §

X) 00

73 i-

JS a 2

"K e« O

£ u «*-

a u C

00 CO "3

o
oo

<u <u

D.-S
U 3

00 o

o —

a
<

wo
cn

>
2

a.
E

N —

'S3
2

2 - a

§ 9J S

■c

3
<

"3
C

^ c

Si
if
>

o
y
c

u
"3

2 *

"o <u

eg
3

t/3

3
C

~ "3 ** fll

_ 3 <U

.S — £ 5

c2 2 2 I 2

" ^

2i o ^
S<i; g

« -3 rf
< Q ^

<u
>

W
oo

— 1

•J

c?5

337

BrandsstaSir : Dry and
good during harvest

ViSvik : Dry and cold. Then
rain set in during harvest
(c.mid August)

Grund : Very cold

Grof : June quite windy
and rainy. Mainly good
and calm from end June
to 14 August. Rainstorm
on 15 August. Calm again
through 18 August. Rain
from 19 August.

SiSumuli : Averagely good

Leira: Cold, Nly with sleet
and night frost to 24 June.
Dry during July and August.
Then became more damp with
Sly winds to mid October.

GarSur : Unpleasant.
Alternating
snow, rain and
frost

KetilsstaSir :
Very damp and
rainy

Reyk javik : On the whole, very
cold and inconstant except for
3-17 August when weather mild

and good

Vik: Heavy rain and storms

11 June to 11 July. 17, 18
and 19 July Skei5ara
flooded. 12 July to 11 August
dry and good. 12 August to

12 September rain

Figure 2: Summer weather reported in Iceland in 1816.

The Sheriff writing at Vidvfle in Skagafjardarsysla stated that the summer was dry and cold
except during the harvest around mid-August when there was rain. Stefan Pdrarinsson at
Modruvellir also mentions long-lasting rain during the harvest. The report from Grund, near
Modruvellir, was of a very cold summer. In the letter from Gardur, the summer is said to have
been unpleasant with alternating snow, rain and frost. The other northern source used here,
Brandsstadaanndll, disagrees with these accounts. According to this source, the weather became
good and calm after 25 April, and the summer as a whole was favourable. It should be noted that
such variability between different areas in Iceland, even sites in close proximity, is not unusual.
Furthermore, Brandsstadir in Blondudalur, where Brandsstadaanndll was written, is in a fairly
sheltered location.

In the east, Sheriff P211 Pordarson Melsted, writing at Ketilsstadir, commented on the damp and
rainy summer. At Vfk, in the southeast, the summer weather was quite variable. However, this
variability took the form of quite long spells of fairly stable weather patterns, rather than short-
term variation on a scale of days. From 1 1 June to 12 July there were storms and heavy rains.
Then, from 12 July to 11 August, there occurred "the driest and best weather of the whole
summer." On 12 August, when the hay harvest had just begun, a rainy period set in and lasted,
with the exception of a very few days, to September. The rainy weather in the southeast in the
early part of the summer may have been partially caused by the volcanic eruption that occurred
in Skaftafellsjokull. This eruption, and the flood in the river Skeidar2, are discussed below.

338

Some western accounts of the summer weather show a pattern of variability not dissimilar to that
noted above for Vflc, although the timing and duration of cold, wet and dry spells is different in
different locations. At Leira' in Borgarfjardarsysla, the Sheriff noted that the weather was cold,
with northeasterly winds, sleet and night frost to about 24 June. Throughout July and August it
was dry. At Reykjavflc, just south of Leira\ the report stated that, on the whole, the weather was
very cold and inconstant except for 3-17 August when the weather was mild and good. The
account from Grof, in Snaefellsnessysla in the west (like Reykjavfk, an exposed coastal site),
contains daily weather data, and is much more detailed than most of the sources used here. It
accords quite well with the Reykjavfk report. The daily data given may be summarized as
follows. On the whole, the month of June was mainly windy with snow, sleet or rain. Ten days
were characterized as being calm. Only one day of rain is mentioned in July, but many days were
described as breezy. No storms occurred in July. From 1-14 August, the weather was quite
favourable. On 15 August and 19-24 August there were rain storms. On 25-27 August there were
strong south-southwesterly winds with rainshowers, and on 28 August the wind was northeasterly
with rain and fog. Northeasterly winds continued to the end of the month. One other western
account, from the inland site of Sfdumuli in Myrasysla, characterized the summer as "averagely
good."

In spite of the reports from Sfdumuli and Brandsstadir, and some intervals of good weather at
other sites, when the summer weather of 1816 is considered over Iceland as a whole, it must be
classed as unfavourable. However, it was not extremely so, and the phrase "a year without a
summer" does not, therefore, seem appropriate for this year in Iceland.

Autumn

Most sources characterize the autumn as mainly stormy and changeable. Thus, in the letter from
Reykjavfk, it is said to have been "stormy all autumn." The report from Grof is of "stormy and
inconstant rainy weather." At Vidvfk, the weather is said to have been wet and inconstant, often
with strong winds. At Leira\ southerly winds are noted. The weather was damp up to mid-
October. At Vfk, September is said to have begun with severe night frost. Subsequently, more
rain than frost occurred. From 3 to 9 October, there was dry good weather with rime frost. From
10 October, there were mainly westerly winds with hail, snow, rain, sleet, frost and layers of ice
on the ground. The autumn weather is said to have been, in general, "unusually changeable."
According to Brands stadaandll, there was snow and frost in late September. The first half of
October was calm and dry, then snows fell.

Other Environmental Events

In 1816, a volcanic eruption also occurred in Iceland. This is known from the letter written by
Sheriff Lydur Gudmundsson at Vfk in Vestur-Skaftafellssysla. According to him, the eruption
began under Skaftafellsjokull (glacier) some time in May. In June, the eruption was visible over
16 miles (24 km) away, with an enormous column of rising vapour. "This later divided itself into
clouds, and caused a bitingly sharp, cold drought until the clouds finally dispersed, and fell as
a malignant, cold, severe and lasting heavy rain." The eruption does not appear to have had any
serious effects on the populace, although the vegetable and hay crops were said to have been
adversely affected.

Lydur Gudmundsson also reported flooding of the River Skeidara" on 17, 18 and 19 July. He
described the river as "flowing out of the bowels of the Skaftafell glacier. " Today, the river flows
adjacent to the neighbouring Skeidararjokull. This discrepancy may be explained by the fact that
the glaciers are undoubtedly smaller and of a different shape now than they were in the early

339

nineteenth century, and the river is also likely to have changed its course. The Sheriff noted that
the river flooded a large part of Skeidarasandur (a stretch of sandy plain, washed out from the
glaciers) and cut off all passage over a much greater distance. Probably the flood was largely
caused by ice melting during the volcanic eruption.

Climatic Impact in Iceland in 1816

The Study of Climatic Impact: Methods and Approaches

In order to provide an analysis of past events that is as accurate as possible, recent research in
the field of climatic impact has emphasized the need to adopt a rigorous methodology (Wigley
et al. 1981; Kates et al. 1985). This is because of the difficulty in isolating and quantifying the
effect that climate might have had on society, given all the other social, political and economic
factors present. Such an exercise is difficult to carry out with present-day data. In the past, when
fewer economic and climatic data were available, it becomes even more problematic. Although
the difficulties arising from this can never be entirely eradicated, a number of measures may be
adopted in order to provide a valid picture of possible climatic impact in the past. Important
issues to consider before undertaking such a study are: (1) the location of the area to be studied;
(2) the quality of the available data; (3) the time scale involved; (4) the economy and social
structure of a given area (e.g., whether primitive or sophisticated); and (5) the strategy, or
methodology to be employed. These points will be considered further below, with regard to
Iceland.

Iceland 's Location

Concerning the first of these points, location, Iceland occupies a marginal area on the borderline
between environments that lend themselves easily to human habitation, and those that do not
(e.g., Bergthtfrsson 1985). The cool climate of the area will obviously be a major factor in
determining the growth of vegetation of all kinds. The many mountain and cold-desert regions
mean that any attempts at agriculture will be limited, not only by the climate, but also by the
amount of land available for such activities. Iceland's geographical situation thus makes it highly
suitable for climatic-impact studies.

Data and Time Scales

Climatic data available for Iceland during the pre-instrumental era, and the specific sources used
here, have been discussed above, and their quality established. Regarding the question of time
scale, the data available make it possible to study climatic impacts in the long term (centuries),
medium term (years to decades), and short term (a year or less) as it is here. The study of
climatic impact in the short term has been criticized as giving undue attention to certain crisis
years (Ingram et al. 1981). However, in this case, it is done within the context of previous
studies of longer periods (Ogilvie 1981; 1984b).

Iceland 's Economy and Society

Before about the mid-nineteenth century, no settlement large enough to be considered a town, or
even a village, existed in Iceland. There were only isolated farmsteads and a few fishing stations
on the coasts. The farms were scattered in order to make best use of the land available.
Settlement was concentrated primarily in the coast and lowland areas of the southwestern, western
and northern regions. The less hospitable areas of the northeast, southeast and northwest were
even more sparsely populated.

340

Iceland's economy was based on animal husbandry: the main animals kept were sheep and cattle.
These provided food in the form of meat and milk products, and also other useful items such as
wool for clothing. Horses were used for transportation. Fishing was also important, but it was
not until the twentieth century that this became a major industry.

During the short summer season, the major task for most Icelanders would be to bring in the
annual hay harvest - still of great importance today. Hay was grown on the "homefields" {tun),
near the farm, and on outlying pastures (engi). The hay was given to livestock during winter so
that they could survive if there was little or no vegetation available. When the weather was
favourable, certain of the livestock, particularly horses, sheep and gelded cattle, were expected
to graze outside. These were known collectively as utigangspeningur or "outside livestock."

From 1380 to the Second World War, Iceland was ruled by Denmark. The trading monopoly
enforced by Denmark for much of this time frequently worked in Iceland's disfavour as the
Danish merchants controlled both prices and the goods available to the Icelanders.

Strategy

Throughout Iceland's recorded history, there are many "crisis-years." These are when the sources
recount failure of the hay crop, livestock deaths, serious difficulties among the population such
as the desertion of farms, begging, and even human mortality. Such events invariably occurred
during very cold years or decades. Because people were so dependent on a successful hay harvest
for supplementary winter fodder, it appears very likely that a poor harvest or a severe winter
might have a considerable impact on the populace.

Rather than take this coincidence of events at face value, however, it is possible to adopt a
strategy that will help to establish more clearly exactly what was occurring. To this end, this
possible impact of climate may be divided into direct and indirect impact.

It is not difficult to demonstrate that climate had a considerable direct impact on biological and
physical processes; (e.g., on grass growth, hay yield, and on other plants). This may be shown
statistically (Ogilvie 1981, 1984b). For example, relationships between temperature, precipitation,
grass growth and hay harvest may be tested by means of contingency tables (e.g., Table 2).

Indirect effects of climatic impact include deaths of livestock by starvation (although such effects
may be compounded by direct impact in the form of cold and damp), and these may also be
demonstrated by means of contingency tables (Table 3). In both of the tables shown here, the
results are highly statistically significant.

Further indirect effects of climatic impact, such as the social problems mentioned above, are far
harder to prove. Yet frequently much circumstantial evidence is available making it possible to
show that such effects were very likely to have occurred (Ogilvie 1981). However, in all such
studies, it is vital to take political, economic and social factors into account, as these invariably
play a larger role than climate.

It is not possible to carry out the kinds of statistical tests mentioned above when considering data
for one year only. However, as climate did have both direct and indirect impacts over longer time
scales, clearly these would also be felt on an annual time scale. The reality of climatic impact has
been demonstrated by several researchers using both modern and historical data (e.g.,
Bergth<5rsson 1966, 1985; Fridriksson 1969, 1972; Bergthtfrsson et al. 1988).

341

Table 2: Summer Temperature and Grass Growth in Iceland 1601-1780'.

Summer Temperature

Grass

growth Cold Average Mild Totals

Poor 52(24.6) 20(22.2) 20(45.2) 92(40.4%)

Average 7(17.1) 9(15.4) 48(31.4) 64(28.1%)

Good 2(19.3) 26(17.4) 44(35.4) 72(31.6%)

Totals 61(25.8%) 55(24.1%) 112(49.1%) 228(100%)

1 chi2 = 84.0

Table 3: Winter Temperature and Livestock Deaths in Northern Iceland 1601-17801.

Winter Temperature

Livestock Cold Average Mild Totals

Deaths 43(20.8) 2(6.5) 2(19.7) 47(28.1%)

Poor condition/disease 5(8.9) 6(2.8) 9(8.4) 20(12.0%)

Good condition 26(44.3) 15(13.8) 59(41.9) 100(59.9%)

Totals 74(44.3%) 23(13.8%) 70(41.9%) 167(100%)

1 chi2 = 62.8

342

In the section below, climatic impact during 1816 is considered. The direct impact of climate on
grass growth and harvest, and on the vegetable crop, plus the direct physical impact of sea ice,
is discussed first. Second, indirect climatic impact on domestic animals and humans is considered.

Direct Impacts of Climate

Grass Growth and Harvest

During 1816, both grass growth and hay harvesting varied considerably around the country in
terms of quality and quantity. Only one source, Brandsstadaann6.ll, from Blondudalur in
Hunavatnssysla, gives an unqualified report that the grass grew well. Haymaking began on 25
July, and there was a successful hay harvest. Another source, that written at Sfdumuli in
Myrasysla, categorizes the grass growth as "quite good", and states that the harvest, and
subsequent use of the hay, also went reasonably well. It is interesting to note that these are the
only two sources which report a good, or, in the latter case, an "averagely good" summer as far
as weather is concerned.

At Ketilsstadir in the east, Sheriff P211 Ptfrdarson Melsted judged that the grass growth was good
"on the whole", although "lack of warmth" meant that the outlying pastures did not grow as well
as the homefields. The harvest, however, was below average. This the Sheriff attributed to the
damp and rainy summer which prevented the hay drying. The Sheriff of Snajfellsnessysla in the
west, Sigurdur Gudlaugsson, who lived at Grof, noted a similar situation. The grass seemed to
grow well, but in the end turned out to be average. "The harvest from the homefields was very
mediocre due to rain and damp weather." Writing about the autumn of 1816, he commented
further: "On account of the autumn's stormy and inconstant rainy weather, the harvest was very
poor in many parts of the district, especially on higher ground where some of the hay blew away
and was washed away from the ground." The opposite situation is reported by J6nas Scheving,
Sheriff of Borgarfjardarsysla at Leir2: "The grass growth, especially in the outlying pastures was
average, but poorer from the homefields. However, the actual harvesting was excellent." The
average to poor grass growth he attributed to cold weather from mid-May to about 24 June and,
more particularly, the dry weather which followed this. The harvest "did not begin until the end
of this month (July)." The dry weather, which lasted to the end of August, undoubtedly facilitated
the harvest. The final state of the grass is also said to have been average in the account written
at Reykjavik. However, grass growth was said to be very late due to cold spring weather. The
harvest was "difficult." A letter of March 1817, states that in Arnes district, in the south, "the
weather is supposed to have been not unfavourable to the harvesting of the hay." Furthermore,
in spite of the difficult harvest, "with the exception of a few individual farms in Kj6s district"
there has not been a lack of hay up to this time. However, the severity of the winter 1816-17
meant that the upland farmers had to give outside livestock hay almost constantly. "It is thus
feared that if the winter should remain severe during the present and next month, the lack of this
item will be considerable."

In most northern districts, the situation regarding grass and hay during the summer and autumn
of 1816 seems to have been more difficult than that of most other regions. Stefan Pdrarinsson,
writing from Modruvellir in Eyjafjardarsysla, commented that, as a result of the cold spring, the
grass growth was no more than average in most places in the north. This is echoed by other
letters from the north. The account from Vidvflc, for example, states that cold, dry weather
prevented grass growth; and at Grund, Sheriff Gunnlaugur Briem noted that grass growth was
unfavourable due to a very cold summer. All these northern letters mention that an epidemic,
which affected people in many parts of Iceland this summer, served to hinder the hay harvest.
The letter from Vidvfk also commented on the rain that set in during the middle of the harvest.
Stefan P6rarinsson also noted that long-lasting rain during September, together with storm winds

343

that blew some of the hay away, caused a setback to the harvest of the outlying pastures, and
resulted in this being, in his opinion, below average.

Comments on grass growth and the harvest in the different sources used here are summarized in
Table 4. Also included is a summary of the characteristics of the winter, spring and summer
seasons. In Table 5 the perceptions of the writers on how the weather affected the grass and
harvest are shown. The main characteristics of the spring and summer seasons, plus grass growth
and hay yield at each location, are shown in Table 6. Spring weather and grass growth are
compared, and summer weather and the harvest. There can be little doubt that the summer
weather directly affected the harvest. For example, if rain or snow or strong winds occurred, the
harvest would be jeopardized. The exact effect of the spring weather on grass growth is much
more complex, involving other variables such as soil condition, use of fertilizer, etc., but, from
previous work (Bergth6rsson 1966; Fridriksson 1972; Ogilvie 1981, 1984b; Bergth6rsson et al.
1988) it is known that unfavourable weather (whether excessively cold, dry or wet) has a
damaging effect on grass growth. It is interesting therefore to compare the incidence of
favourable/unfavourable weather with favourable/unfavourable grass growth or harvest in the
different locations (Table 6). Where these coincide a line is drawn between them. The harvest
and summer weather agree in every case but one. However, it would be reasonable to assume
agreement in this latter case also, as the Modruvellir site, where the summer weather was not
reported, lies only a few kilometres from Grund where the weather was said to be very cold.
Grass and spring weather, as might be expected, do not agree as well, but the agreement (in six
out of 11 cases) is nevertheless striking.

Vegetable Cultivation

From the latter part of the eighteenth century onwards, a serious attempt was made by the Danish
authorities, and by enlightened individuals, to get ordinary people to supplement their diet by
growing vegetables. The most commonly-planted species were potatoes, cabbage and turnips.
These crops failed almost everywhere in Iceland in 1816. At Vidvfk in the north, for example,
Sheriff J6n Espblm noted that the number of gardens in use had increased greatly, but that they
had not done well this year due to "the severe weather and storms" and also to the epidemic
which affected people almost the whole summer, and prevented them from working. Early in
1817, he wrote again, commenting that gardening activity had ceased as the ground was frozen.
He continued: "... one cannot think without sorrow of... the many years of dearth in most places
in this district..."

Accounts from elsewhere for 1816 are similar to J6n Espdlm's. Stefan Ptfrarinsson, writing from
Modruvellir, stated that some turnips and cabbage had grown, but that the potato harvest had
failed completely. Sheriff J6nas Scheving, at Leira\ wrote that vegetables had done very badly
over the past year. This he attributed to lack of sufficient seed, and also to cold spring weather,
and dry weather in July. A poor vegetable crop also occurred in Vestur-Skaftafellssysla.

However, the Sheriff there, Lydur Gudmundsson, mainly attributed their "pale and sickly
appearance" to the effects of the volcanic eruption that occurred under Skaftafellsjokull in June
1816.

As with the grass growth and harvest, it seems reasonable to assume that, aside from the effects
of this eruption, the weather of 1816 did play a considerable role in the failure of the vegetables.
This is also suggested by previous work on crop/climatic relationships (e.g., Parry et al. 1988).

344

The Impact of Sea Ice

As noted in the early part of this paper, Iceland is close to the seasonal boundary of Arctic drift
ice. When the ice reaches Iceland (most commonly, the northern, northwestern and eastern
coasts) the most striking climatic effect is a lowering of temperatures in the areas affected (see
also Wilson, this volume, regarding the cooling effect of sea ice lingering near the eastern coast
of Hudson Bay). Rain and mist may be associated with the ice. The presence of the ice also has
a direct physical impact. Because the ice prevents access to the open sea or makes it hazardous,
activities such as fishing and sealing are prevented or hindered. This is no less true today than
in past centuries, but, in the twentieth century, sea ice has not been common near Iceland. Other
activities, such as gathering of shellfish from the shore, and the grazing by livestock of seaweed
and marine plants, are also curtailed by land-fast ice. Such dietary supplements for humans and
animals are of relatively little importance today, but played a vital role in the past.

The sea ice did bring some benefits, mainly in the form of driftwood and the occasional beached
whale or other sea mammal, driven ashore by the encroaching ice. Wood was always in short
supply, and a whale would greatly augment the food supply. For a more detailed discussion of
the effects of the sea ice on flora and fauna, see Fridriksson (1969).

During the period 1809-20, heavy ice years occurred in 1811, 1812 and 1817. During these
years, ice was present off the northern coasts and elsewhere from some time in January to July
or August. During 1818 and 1819 very little ice appeared. The former year was very unusual in
that the sea ice occurred in August, although not for long. In the latter year ice was seen briefly
in April.

The year 1816 may be classed as a moderate ice year. During this year, sea ice affected the
northern coast of Iceland from the beginning of March to the middle of July. Stefan Pdrarinsson,
at Modruvellir, commented that the ice caused persistent northerly winds, frost and cold air.
Briefly, ice prevented the arrival of the first trading ships at Eyjafjord. At Gardur, in Sudur-
Pingeyjarsysla, Sheriff Pordur Bjornsson stated that the seal fishing had been very good until sea
ice came and prevented this. The shark fishing was poor for the same reason. According to the
account at Grdf, ice also prevented fishing in parts of Breidafjordur, in the west. But the layers
of ice "far out to sea" reported by Sigurdur Gudlaugsson, were caused by the sea itself being
frozen, and not by actual sea ice. The Sheriff commented that in the 1 1 years he had been there,
the fishing had never been as poor as this year.

Although sea ice undoubtedly caused some inconvenience during 1816, there is little evidence to
suggest that it had a major impact on food supplies.

Indirect Impacts of Climate

Livestock

Most sources mention the severe winter this year, and the frequently frozen ground that prevented
grazing. Nevertheless, there were no serious losses of livestock. Indeed, only Stefan Pdrarinsson,
writing at Modruvellir in Eyjafjord district in the north, reported that some people lost a number
of their outside livestock. He wrote:

345

Table 4: Summary of Seasons, Grass Growth and Harvest in 1816.

Place

Seasons'

Grass

Harvest

Ketilsstadir

Gardur

Modruvellir

W - Average
Sp - Cold, calm
Sm - Wet

W - Very severe
Sp - Sea ice present
Sm - Rain, snow, frost

W - Very severe
Sp - Quite severe
Sm -

Good on homefields; Below average
not as good on out-
lying pastures

Poor

Average

Below average

Grund

Vidvflc

Brandsstadir

W -

Sp -

Sm - Very cold

W - Severe

Sp - Dry and cold

Sm - Dry & cold then wet

W - Mainly severe
Sp - Weather improved
Sm - Dry and good during
harvest

Unfavourable

Poor
Good

Unfavourable

Poor

Good

Grof

W - Fairly severe
Sp - Variable
Sm - Variable

Average

Very poor

Sidumuli

Leira

Reykjavik

Vile

W - Severe
Sp-

Sm - Averagely good

W - Very severe
Sp - Mainly severe
Sm - Dry to end Aug.

W - Severe
Sp - Unusually cold
Sm - Mainly cold and
inconstant

W - Very severe

Sp - Severe

Sm - Unfavourable

Quite good

Reasonably good

Average on pastures; Very good
poorer on homefields

Average

Difficult

Poor

Meagre and spoilt

1 W winter; Sp spring; Sm summer

346

Table 5: Contemporary Perceptions of Climatic Impact on Grass Growth and Harvest in 1816.

Ketilsstadir

Cold spring meant that the outlying pastures did not grow as well as the homefields. Nevertheless, grass
growth good on the whole. Harvest below average due to wet summer. Not possible to dry hay - therefore
stacked up damp.

Gardur

Harvest poor due to bad weather and epidemic.
Modruvellir

In spite of the cold spring, the grass growth was about or almost average in most places here in the north.
In the east it is said to have been poorer. The summer's harvest did not live up to the promise of the grass
growth, however. This was due to the epidemic which occurred everywhere in the north at the beginning
of the harvest. Then rains in September plus storm winds adversely affected the hay on outlying pastures.
Thus, on the whole, harvest below average.

Grand

Grass growth unfavourable due to cold summer. Harvest also, primarily due to epidemic.
Vidvik

Grass did not grow well due to cold spring and summer weather. Harvest poor due to rains and epidemic.
Grof

In most places the grass growth looked quite good to begin with, but turned out to be only average and,
on account of wet weather, the harvest of the homefields was mediocre. Due to stormy and inconstant rainy
weather, harvest very poor in many parts of the district, especially on higher ground where some of the
hay blew away it was washed away from the ground.

Sidumuli

Dry weather from about 24 June meant that grass growth poorer than last years, so harvest did not begin
until end July.

Reykjavik

As a result of the cold spring weather, the grass growth was only average and the harvest very difficult.
Nevertheless, most people do not lack hay.

Vik

Poor grass growth due to volcanic eruption. Harvest spoiled by rains.

347

<U
60
CO
u.

>
13

O
O
00

X)
CO

"8.

u
60

o
o

CO
O
-*— »
3
£>

"o
O

fed

u
CO

4>
>

o
O

1
o

60
a)
u.

<0
>
<

'«

42
CO

O
>
hh

O
>

(4-1

o
o

-a
"o

60
CO
w.

U
>
<

•a

o
o

•u
60
c«
u.

>
<

60
at
Ih

•8
>

O
I

•8
■c

CO
>

D
CO

2
o

^3
E
3

U
fc

CO
43

>>
CO
43
Ih
O

J3
%

oo
2

O
>

y
c
u

T3
.S

348

... the long lasting layers of ice in most places in this region... caused a good many
farmers, here and there, to suffer a lack of fodder. They therefore lost a number of
their so-called outside livestock, especially horses, due to emaciation. However, the
latter loss (of the horses) only applied to some of the inhabitants of Skagafjord and
Hunavatn districts who, to their own detriment, keep far too many horses. On the
whole, the loss of outside livestock was neither general, nor of great importance.

Two other letters reported that lack of fodder meant that some livestock had to be slaughtered.
These letters are from Gardur, in Adaldalur in the north, and Grof in Snaefellsnessysla, in the
west. The account from the former stated that, although some people were forced to slaughter
their livestock toward the spring, livestock deaths were not general. The latter source commented
that in many places people had to slaughter their sheep as the usual winter grass failed in most
places.

Other sources remark on the difficulties for livestock during 1816, but emphasize that, on the
whole, they were kept alive. At Ketilsstadir, in the east, the winter was said to be only average
but, because of large amounts of snow in some places, the outside livestock had to be given
fodder for a long time. However, "this did not last so long that the animals died of hunger."
Sheriff Peftir Otteson, writing from Sfdumuli in the west, stated that, because of the layers of ice
and snow, virtually no grass was available for the livestock. He continued: "they would have died
in great numbers if there had not been sufficient fodder after last year's good harvest.

The letter from Vflc does not comment on the livestock during the winter, but says that, during
heavy rain and storms from 1 1 June to 1 1 July, cows and ewes needed food and shelter. The
Sheriff added: "After the severe winter, this could scarcely be spared," implying that, here too,
the livestock needed extra fodder during the winter. According to this letter, the poor hay harvest
this year caused livestock, especially cows, to be slaughtered in the autumn. From 9 October
onwards, changeable weather with "hail, snow, layers of ice, rain, sleet and frost" meant that
little grass was available. The Sheriff at Vflc commented: "The horses and sheep have become
emaciated and have sometimes needed to be given fodder, and this has had to be shared with the
few remaining cows." The state of the livestock in the autumn and early winter is also noted by
the Sheriff of Borgarfjardarsysla, at Leira\ After mid-October, "the winter set in with alternating
frost and drifting snow, thaws and rain. This made it very difficult for the livestock, the horses
and sheep who need to find their own food, as the frost caused the large quantities of snow and
water which fell to form a frozen layer on the ground." We may conclude that, during 1816,
conditions for livestock were, if not easy, not usually difficult either.

Social Stress

Research carried out for the period 1601 to 1780 (Ogilvie 1981) has shown that it is very likely
that during this time climate did play a part in the occurrence of social stress, which manifested
itself in such phenomena as the desertion of farms, begging and petty crime, plus hunger-related
diseases and mortality among the people. During 1816, however, such problems were not
widespread. Only one district reported general difficulties of this kind. This was Snaefellsnessysla,
in the west. Here, Sheriff Sigurdur Gudlaugsson wrote:

Great lack of food among inhabitants. People pressed by beggars from here and also
from other districts. The majority of the district's populace have already got into debt
at the trading places in previous years, and have scraped together all that they could
in order to pay. So now they have to give all the best fish to the merchants and have
little left for themselves except for flatfish and cod's heads. This is poor winter

349

provision, particularly on the coast among the poor fishermen who do not earn
sufficient during the summer to huy other necessary foodstuffs from the farmers, and
who therefore frequently live in the greatest misery.

The lack of food must be partly attributed to the fact that, as the Sheriff noted elsewhere in his
letter, the trading places were very poorly supplied with corn wares and other imported
foodstuffs. Furthermore, the fishing, of great importance in this district, largely failed this year.
Clearly this was largely due to climate. The Sheriff describes how "although there should have
been fishing in the latter part of the winter months, the severe frost and layers of ice far out to
sea, frequently prevented the fisherman from getting out to sea for many days on end."

Because Snaefellsnes and nearby areas were important fishing centres, they attracted people whose
inland sources of food had dwindled. Thus, although most other districts do not report social
stresses this year, their silence on such matters may be partly attributable to the fact that the
people in difficulties had already left to try their luck at the western and southern fishing stations.

Conclusions

During 1816, most districts in Iceland experienced a very severe winter. One source, the letter
written by Sheriff J6n Esp61m at Vidvfk, compared it with two other very severe winters in
recent times, 1784 (see Wood, this volume, regarding climatic effects of the Laki eruption) and
1802. The spring was also mainly severe in most places. It was a moderate sea-ice year, with ice
present off the northern coasts from the beginning of March to mid-July. The summer was
unfavourable, at least for part of the season, in most districts in Iceland with various
combinations of excessive cold, wet or drought reported. In certain parts, the epithet "year
without a summer" may have been appropriate, but if we consider the whole summer, over all
Iceland, then it would not have been. The regional variability reflected in the sources used here
is quite in accord with what is known of local climatic effects in Iceland (Eythdrsson and
Sigtryggsson 1971; Ogilvie 1984a).

If the summer of 1816 had been unfavourable in all parts of Iceland, as happened in true "years
without summers" such as 1756 (Ogilvie 1981) and 1783 (Ogilvie 1986), then the climatic impact
felt might have been greater. However, it might also have been greater if a favourable harvest
had not occurred in 1815, thus boosting haystocks.

It is not difficult to demonstrate that direct impact, for example, on grass growth and hay yield
did occur in 1816. The indirect role of climate on society this year is harder to define. While it
is clear that there were difficulties amongst the populace, these were not widespread and were
compounded by political and economic factors (e.g., by difficulties with trade). Several accounts
this year report that supplementary foodstuffs received from Denmark were insufficient or of poor
quality. There were also reports of poor fishing catches. It is true that fish are affected by
climate, but the relationship is complex and, as yet, not fully documented. Certainly, poor fishing
catches at sea are not directly linked to climate on land except in the case of heavy storms or
when lowered temperatures cause ice to form on the sea, thus preventing fishing (as occurred off
Snaefellsnes district this year). The presence of sea ice may also hinder fishing as happened off
the north coast of Iceland this year.

350

In spite of the difficulty in allotting specific roles to economic, political and climatic factors in
the general well-being of the Icelanders in 1816, there can be little doubt that some indirect
climatic impact was felt this year. In the climatic context alone, 1816 was certainly an interesting
year, if not a "year without a summer."

Acknowledgements

Dick Harington, Tim Ball and Cynthia Wilson deserve praise for their efforts in organizing the
meeting " The Year Without a Summer? Climate in 1816" held in Ottawa June 1988. As always,
I am grateful to many Icelanders for their help. Here I should like to acknowledge in particular
P6rhallur Vilmundarson, Adalgeir Kristjansson and Trausti J6nsson. Part of the research for this
paper was supported by grant GR3/7013 from the Natural Environment Research Council. This
paper is dedicated to Valmore C. La Marche Jr. (1937-1988), who had been looking forward to
joining in the debate on the climate of 1816.

I had a dream, which was not all a dream
The bright sun was extinguish'd, and the stars
Did wander darkling in the eternal space,
Rayless, and pathless, and the icy Earth
Swung blind and blackening in the moonless air.

(From "Darkness" by Lord Byron. Written in 1816.)

References
Manuscript Sources

Pj6dskjalasafn (National Archives), Reykjavik.
Islands Journal 12

Stefan Pcjrarinsson. Modruvellir. 26 September 1816, no. 2621.

Gunnlaugur Briem, Grund. 5 October 1816, no. 2650.

J6n Esp61m, Vidvfk. 26 August 1816, no. 2593 and 30 June 1817, no. 2818.

Sigurdur Gudlaugsson, Grof. 18 February 1817, no. 2811.

Pemr Ottesen, Sfdumuli. 31 December 1816, no. 2808.

Johan Carl Thueracht v. Castenskiold, Reykjavik. 17 August 1816, no. 2519 and 5 March
1817, no. 2847.

Islands Journal 13

Pall P6rdarson Melsted, Ketilsstadir. 19 October 1816, no. 392 (formerly Islands Journal 12,
no. 2645).

J6nas Scheving, Leirl 31 July and 31 December 1816, no. 24.

351

Lydur Gudmundsson, Vflc. 17 January 1817, no. 27.
Islenzka Stjdrnardeild 8

Pordur Bjornsson, Gardur. 23 September 1816, no. 505 (formerly Islands Journal 12, no.
2644).

Published Sources

Bell, W.T. and A.E.J. Ogilvie. 1978. Weather compilations as a source of data for the
reconstruction of European climate during the medieval period. Climatic Change 1:331-
348.

Bergthdrsson, P. 1966. Hitafar og busaeld a Islandi. Vedrid 11(1): 15-20.

. 1969. An estimate of drift ice and temperature in Iceland in 1,000 years. JOkull 19:94-

" 101.

. 1985. Sensitivity of Icelandic agriculture to climatic variations. Climatic Change

7:111-127.

Bergthtfrsson, P., H. Bjornsson, O. Dyrmundsson, B. Gudmundsson, A. Helgadtfttir and
J.V. Jtfnmundsson. 1988. The effects of climatic variations on agriculture in Iceland. In:
The Impact of Climatic Variations on Agriculture. Volume 1: Assessment in Cool
Temperate and Cold Regions. M.L. Parry, T.R. Carter and N.T. Konijn (eds.). Kluwer
Academic Publishers, Dordrecht, Boston, London, pp. 383-509.

Bjarnason, B. 1941. Brandsstadaanndll . S6guf61agid Hunvetningafdlagid f Reykjavfk,
Reykjavflc. 237 pp.

Bradley, R.S. 1988. The explosive volcanic eruption signal in northern hemisphere continental
temperature records. Climatic Change 12:221-243.

Einarsson, M.A. 1976. Vedurfar d Islandi. Idunn, Reykjavik. 150 pp.

EypCrsson, J. 1956. Elztu Vedurathuganir med maelitaekjum £ Islandi. Vedrid l(l):27-28.

Eythdrsson, J. and H. Sigtryggsson. 1971. The Climate and Weather of Iceland. In: The
Zoology of Iceland 1(3). S.L. Tuxen (managing ed.). Ejnar Munksgaard, Copenhagen and
Reykjavik, pp. 1-62.

Fridriksson, S. 1969. The effects of sea ice on flora, fauna and agriculture. Jdkull 19:146-
157.

. 1972. Grass and grass utilization in Iceland. Ecology 53:785-796.

Gunnlaugsson, G.A., G.M. Gudbergsson, S. Porarinsson, S. Rafnsson and P. Einarsson.
(Editors). 1984. Skaftdreldar 1783-1784 Ritgerdir og Heimildir. Mai og Menning,
Reykjavfk. 442 pp.

352

Ingram, M.J., D.J. Underhill and T.M.L. Wigley. 1978. Historical climatology. Nature
276:329-334.

Ingram, M.J., G. Farmer and T.M.L. Wigley. 1981. Past climates and their impact on Man:
a review. In: Climate and History. Studies in Past Climates and Their Impact on Man.
T.M.L. Wigley, M.J. Ingram and G. Farmer (eds.). Cambridge University Press,
Cambridge, pp. 3-50.

Kates, R.W., J.H. Ausubel and M. Berberian. (Editors). 1985. Climatic Impact Assessment.
Studies of the Interaction of Climate and Society. SCOPE 27. Scientific Committee on
Problems of the Environment (SCOPE) of the International Council of Scientific Unions
(ICSU). John Wiley and Sons, Chichester, New York, Brisbane, Toronto, Singapore. 625
pp.

Kelly, P.M. and C.B. Sear. 1984. Climatic impact of explosive volcanic eruptions. Nature
311:740-743.

Kelly, P.M., T.M.L. Wigley and P.D. Jones. 1984. European pressure maps for 1815-16, the
time of the eruption of Tambora. Climate Monitor 13:76-91.

The King's Mirror (Speculum Regale - Konungs Skuggsjd). 1917. Translated by
L.M. Larson. Scandinavian Monographs!). The American-Scandinavian Foundation, New
York. 388 pp.

Kington, J. A. 1972. Meteorological observing in Scandinavia and Iceland during the
eighteenth century. Weather 27:222-233.

Koch, L. 1945. The East Greenland ice. Meddelelser om Gronland 130(3):\-373.

Lamb, H.H. 1970. Volcanic dust in the atmosphere; with a chronology and assessment of its
meteorological significance. Philosophical Transactions of the Royal Society of London
A266:425-533.

Landsberg, H.E. and J.M. Albert. 1974. The summer of 1816 and volcanism. Weatherwise
27:63-66.

Ogilvie, A.E.J. 1981. Climate and society in Iceland from the medieval period to the late
eighteenth century. Unpublished Ph.D. dissertation. School of Environmental Sciences,
University of East Anglia, Norwich. 504 pp.

. 1984a. The past climate and sea-ice record from Iceland, Part 1: data to A.D. 1780.

Climatic Change 6:131-152.

. 1984b. The impact of climate on grass growth and hay yield in Iceland: A.D. 1601 to

1780. In: Climatic Changes on a Yearly to Millenial Basis. N.A. Morner and W. Karlen
(eds.). D. Reidel Publishing Company, Dordrecht, pp. 343-352.

. 1986. The climate of Iceland, 1701-1784. JOkull 36:57-73.

353

. 1990. Documentary evidence for changes in climate in Iceland A.D. 1500 to 1800. In:

Climate Since A.D. 1500. R.S. Bradley and P.D. Jones (eds.). Harper Collins, (in press).

. 1991. Climatic changes in Iceland c.A.D. 865 to 1598. Acta Archaeologica. (in press).

Parry, M.L., T.R. Carter and N.T. Konijn. (Editors). 1988. The Impact of Climatic Variations
on Agriculture. Volume 1: Assessment in Cool Temperature and Cold Regions. Kluwer
Academic Publishers, Dordrecht, Boston, London. 876 pp.

Sear, C.B., P.M. Kelly, P.D. Jones and CM. Goodess. 1987. Global surface-temperature
responses to major volcanic eruptions. Nature 330:365-367.

Sigfusdtfttir, A.B. 1969. Hitabreytingar & Islandi 1846-1968. In: Haffsinn.
M.A. Einarsson (ed.). Almenna B6kaf61agid, Reykjavflc. pp. 70-79.

Stommel, H. and E. Stommel. 1979. The year without a summer. Scientific American
240(6): 134-140.

Stothers, R.B. 1984. The great Tambora eruption in 1815 and its aftermath. Science
242:1191-1198.

Vilmundarson, P. 1972. Evaluation of historical sources on sea ice near Iceland. In: Sea Ice:
Proceedings of an International Conference, 10-13 May 1971 . P. Karlsson (ed.). National
Research Council, Reykjavflc. pp. 159-169.

Wigley, T.M.L., M.J. Ingram and G. Farmer. (Editors). 1981. Climate and History. Studies in
Past Climates and Their Impact on Man. Cambridge University Press, Cambridge.
530 pp.

354

First Essay at Reconstructing the General Atmospheric Circulation
in 1816 and the Early Nineteenth Century

H.H. Lamb1

Reconstructions of the general atmospheric circulation in January and July year by year back to
1750, based on the best available network of monthly mean M.S.L. barometric pressure values
over as much of the world as possible, from observation data in the archives and library of the
United Kingdom Meteorological Office, were published by Lamb and Johnson (1959, 1961,
1966). The maps were all analyzed by me, and the analyses were tested by a simulation
procedure: maps of the years 1919-39 were first analyzed using only restricted networks of data
corresponding to the information available in the period 1786 to 1820, and these were then
compared with maps for the same (inter-war) years analyzed with the use of full data. The
distribution of errors was then studied. On this basis, it was decided that isobars on maps drawn
for years in the late eighteenth and early nineteenth centuries could be considered satisfactorily
reliable within regions where the standard error on the test maps was less than 1.0 mb in July
(or less than 2.5 mb in January, this figure corresponding approximately to the ratio of the
standard deviation of the observed values in January compared with those in July).

This meant in practice that isobars could only be presented with confidence over, or very close
to, Europe between southern Scandinavia, Britain and the western Mediterranean on the maps for
individual Januarys and Julys in the decade 1810-19. Decade and longer-term mean isobars could
be reliable over a wider area, spanning most of the Atlantic Ocean between latitudes about 30 and
50 to 65°N. Isobars at 5-mb intervals were printed as unbroken lines over the areas established
by the tests as reliable within the limits mentioned (and in regions of slack pressure gradients an
intermediate isobar might be drawn in at a 2. 5-mb interval).

On the maps for individual Januarys and Julys the isobars were extended, as broken lines, over
regions where it seemed that the pattern must be broadly reliable, though the pressure values
could not be relied upon.

In the case of July 1816 - as with some other seasons of historically dramatic weather - use could
be made of a wealth of descriptive data on the weather experienced in many places so that it
seemed reasonable to extend the isobar pattern, as broken lines, far beyond the limits of where
the pressure values were known. This produced the map for the average conditions prevailing in
July 1816 (Figure 1).

The coldness of that summer in eastern Canada, and in northeastern North America generally,
appears here as attributable to prevalence of air drawn directly from the Canadian Arctic and the
closeness of a focus of cyclonic activity to Labrador, Newfoundland and off-lying waters. The
coldness of the summer in Britain, southern Scandinavia and the western part of continental
Europe is seen to be due to the prevailing concentration of a low pressure region - unusually far
south for summer - over the areas named, together with indraught of Arctic air from the source
regions nearby. This is a similar explanation to that more tentatively shown for northeastern
North America.

1 Climatic Research Unit, University of East Anglia, Norwich NR4 7TJ, U.K.

355

Figure 1: Average pressure conditions reconstructed for the area between eastern North America and
western Europe (July 1816). See text for explanation.

356

The much better weather (and crops) experienced in Shetland - and to some extent all over the
northern half of Scotland - and elsewhere in northern and also eastern Europe, extending south
to the Crimea, is readily attributable to the higher pressures (and probably greater sunshine) over
those areas.

References

Lamb, H.H. and A.I. Johnson. 1959. Climatic variation and observed changes in the general
circulation: Parts I and II. Geografiska Annaler 41:94-134.

. 1961. Climatic variation and observed changes in the general circulation: Part III.

Geografiska Annaler 43:363-400.

. 1966. Secular variations of the atmospheric circulation since 1750. Geophysical

Memoir 110. (Her Majesty's Stationery Office, for Meteorological Office). London.
125 pp.

357

Weather Patterns over Europe in 1816

John Kington1
Abstract

An outline of the state of meteorology during the early nineteenth century is presented with
particular reference to the introduction of the synoptic method of analyzing daily weather maps
by Heinrich Brandes in 1816.

Links with the historical weather data made and collected in the 1780s are mentioned in relation
to the series of daily weather maps for Europe that I am constructing from 1781.

The feasibility of undertaking a program of similar research for a period of years centred on 1816
is discussed. As an example, a run of daily charts for Europe in July 1816 is presented together
with a preliminary analysis of the circulation patterns brought to light in the process.

Comparisons are made with events in the 1780s, in particular the cold summer of 1784 that
followed the formation of the exceptional volcanic dust veil after the great eruptions in Iceland
and Japan the preceding year.

Historical Weather Data: Comparison of 1816 with the 1780s

Writing in Breslau, Silesia towards the close of 1816, the German meteorologist Heinrich Brandes
observed:

... If one could collect very accurate meteorological observations, even if only
for the whole of Europe, it would surely yield very instructive results. If one
could prepare weather maps of Europe for each of the 365 days of the year,
then it would be possible to determine, for instance, the boundary of the great
rain-bearing clouds, which in July [1816] covered the whole of Germany and
France; it would show whether this limit gradually shifted farther towards the
north or whether fresh thunderstorms suddenly formed over several degrees
of longitude and latitude and spread over entire countries ... In order to
initiate a representation according to this idea, one must have observations
from 40 to 50 places scattered from the Pyrenees to the Urals. Although this
would still leave very many points uncertain, yet by this procedure, something
would be achieved, which up to now is completely new.

As a meteorological observation network did not then exist, Brandes was unable to examine the
weather conditions of July 1816 but pursued his hypothesis by making use of data collected
30 years earlier by the Societas Meteorologica Palatina. Thus the first observations to be studied
by means of the synoptic method devised by Brandes were those for 6 March 1783 (Figure 1),
a day on which, like many of those in July 1816, stormy weather prevailed over western and
central Europe.

Climatic Research Unit, University of East Anglia, Norwich NR4 7TJ, U.K.

358

Figure 1: Synoptic weather map for 6 March 1783 by H.W. Brandes, reconstructed by H. Hildebrandsson.

Surface wind directions are shown by arrows and the field of pressure by isopleths of equal
departure of pressure from normal (e.g., -17, -16, -15, etc.)- By overcoming the uncertainty
about the height at which the barometer readings were made, the observations were successfully
combined to allow the equivalent of isobars to be drawn at a constant level (from Ludlam 1966).

359

During the Enlightenment, hopes had been raised in scientific circles that a systematic study of
meteorological observations would show that the seemingly disordered array of weather variations
were subject to predictable forms of behaviour. Consequently, extensive networks of observing
stations were established by two scientific societies in Europe during the early 1780s, namely,
the Sociiti Royale de Mtdecine and the Societas Meteorologica Palatina centred at Paris and
Mannheim, respectively.

Unfortunately, tangible results proved to be elusive by the statistical approach then applied, which
earlier had been so successful in predicting the motion of the stars and planets. However, the
collections of reports from these two societies, together with further data from private individuals
and ships' logs, means that a large array of daily instrumental and quantitative observational
material became, and is still, available for the 1780s over Europe. These data provide a network
of more than the "40 to 50 places" advocated by Brandes (Figure 2). This information is now
being subjected to twentieth-century concepts in synoptic meteorology to "yield very instructive
results" in the construction of daily historical weather maps as envisaged by Brandes 170 years
ago. As an example of the series of charts now becoming available from 1781, the map for the
same day as earlier constructed by Brandes (6 March 1783) is illustrated in Figure 3.

The early blossoming of meteorology in the late eighteenth century was brought to a halt by the
political confusion and social unrest that followed the outbreak of the French Revolution. The two
main scientific societies which had been promoting international cooperation in the exchange of
weather information were disbanded in the mid-1790s. After a lapse of two decades, was it the
"year without a summer" with its exceptionally cold wet weather and disastrous harvests that
provided the stimulus for a revival of efforts to understand and predict weather changes? In any
event, the idea of mapping over a large area simultaneous daily observations of meteorological
elements such as pressure, wind and temperature (the concept upon which synoptic weather
studies are based) was, as earlier stated, presented at this time by Brandes.

Having demonstrated recently that it is indeed possible to map historical weather data on a daily
basis over Europe for the 1780s (Kington 1988), and knowing that comparable, albeit less well
sorted and organized, observations are available for 1816, a pilot scheme was initiated for a
monthly period in that year, following the kind invitation to attend this conference by Dr. C.R.
Harington. July was chosen for several reasons, not least being the month first highlighted by
Brandes in his letter of 1816, as quoted above.

In 1967 the German climatologist, Hans Von Rudloff, examined the weather patterns of 1816 in
his study of the fluctuations and oscillations of European climate since the beginning of
instrumental weather observing in the seventeenth century. His analysis showed that there was
an abnormal distribution of pressure over Europe in the summer of 1816. The subtropical high
pressure system, the "Azores High", which usually extends northeastwards over the region at
times during the summer, appears to have been completely absent. Instead, systems of low
pressure persisting over central Europe allowed polar air streams to be advected farther south
than normal over the region (Von Rudloff 1967).

At about the same time as Von Rudloff s study, Hubert Lamb presented an investigation of
secular variations in atmospheric circulation since 1750 by means of a series of maps showing
mean pressure distribution for the months of January and July (Lamb 1967). The chart for July
1816 (Lamb, this volume) again shows an unusual distribution of pressure, with the "Icelandic
Low" positioned well to the south of its normal summer latitude.

360

Figure 2: Map of stations showing the synoptic coverage available for the 1780s (from Kington 1988).

361

Figure 3: Synoptic weather map for 6 March 1783 (from Kington 1988).

362

Figure 4: Synoptic weather map for 7 July 1816 illustrating the Lamb British Isles Cyclonic weather type.

363

Figure 5: Synoptic weather map for 27 July 1816 illustrating the Lamb British Isles North Westerly
weather type.

364

More recently, in the reconstructions of monthly pressure patterns for Europe back to 1780 based
on principal components regression techniques, Jones, Wigley and Briffa presented a map of
pressure anomalies for July 1816 that shows an unusually large negative area, in excess of seven
millibars, over the British Isles and southern North Sea (Jones et al. 1987).

All these works strongly indicate that some very pronounced regional anomalies occurred in the
circulation over Europe in July 1816. Can we discover more? Yes, because an investigation of
weather patterns on a daily basis can reveal aspects of atmospheric behaviour that are not possible
to detect from studies made on monthly or longer time scales.

Although, as previously stated, the two major observation networks of the 1780s were disbanded
in the following decade, a number of the original stations continued in operation, while others
were newly established during the early part of the nineteenth century. Using a nucleus of such
data (readily on hand in the Climatic Research Unit), a run of daily weather maps for July 1816
was specially prepared for this book.

The charts have been analyzed and classified with Professor Lamb according to his system of
British Isles weather types (Lamb 1972). This scheme aims to represent the main types of
circulation patterns prevalent over the British Isles, namely: Westerly (W), North Westerly (NW),
Northerly (N), Easterly (E), Southerly (S), Anticyclonic (A) and Cyclonic (C). Since the British
Isles are centrally placed in the mid-latitude westerly wind belt, as well as being located in one
of the sectors around the northern hemisphere most frequently affected by blocking of this flow,
variations in the circulation over the more extensive North Atlantic-European region are also well
registered by this classification. The classification for July 1816 is given in Table 1.

A statistical analysis of the classification (Table 2) shows that the circulation over the British Isles
during July 1816 was strongly dominated by Cyclonic weather types (three times more frequent
than usual). Of the other patterns, Northwesterly types were also more prevalent than usual (over
twice as frequent); Southerly types about average; Westerly types, however, were about one third
of the normal frequency, while Anticyclonic, Northerly and Easterly types were totally absent.

Typical examples of the two predominant weather types, Cyclonic and Northwesterly, are shown
in Figures 4 and 5.

In Table 3 the frequencies of the Lamb British Isles weather types in July 1816 are compared
with those for 1868-1967, 1781-85 and 1785.

This shows that frequency values for July 1816 are nearer to the averages for 1781-85 (a period
in the Little Ice Age) than those of the standard period, 1868-1967. In particular, the circulation
of 1816 closely parallels that of 1785 with its notable increases in Cyclonic and North Westerly
types and corresponding decreases in Anticyclonic and Westerly types.

The Lamb British Isles weather types are also used to determine the PSCM indices of:
progression, meridionality and cyclonicity, which provide a ready means of indicating the general
character of the circulation over the region for a period of a month or more (Murray and Lewis
1966).

365

Table 1: Lamb British Isles Weather Types for July 1816.'

CNW

1 C Cyclonic, CNW Cyclonic North Westerly, NW North Westerly, S Southerly, SW South Westerly, CS Cyclonic
Southerly, W Westerly, U Unclassified.

Table 2: Lamb British Isles Weather Types. Monthly Frequencies for July 1816 with Long-Period Mean
Percentage Values Given in Brackets for Comparison.1

Days

Vh 1

(26)

lA UV/2 1

(7)

(4)

1 lA Vi

(5)

(24)

1 1 1 1 1 1 1 1 1 1 xh lining \ vh

20VS

(22)

(5)

1 W Westerly, NW North Westerly, N Northerly, E Easterly, S Southerly, A Anticyclonic, C Cyclonic, U
Unclassified.

In July 1816: P = -12 or -3; S = -3 or -1; C=+41 or +42 and M= 11 or 131
That is: P12 Sj4

This shows that the circulation over the British Isles in July 1816 was characterized by blocked
or quasi-stationary cyclonic weather systems. The C index value of +41 or +42 is far greater
than the maximum value of +30 (1936) in the official long-period record from 1861.
Interestingly this record was also broken in the 1780s when the C index in July 1785 was +33.

In July 1785: P = -3; S = -14; C= +33; and M= 14
That is: P2 S, C5 M3

The slightly differing results are due to dealing with a run of charts from a single isolated month, resulting in a
certain lack of synoptic continuity at the beginning and end of the series.

366

Figure 6: Rainfall anomalies (%) for July 1816.

367

Table 3: Lamb British Isles Weather Types. July Frequencies for 1816 and 1785; Period Average
Frequencies for 1868-1967 and 1781-85.'

Number of Days

1816

2.5

5.0

0.0

2.0

0.0

20.5

1868-1967

8.1

2.2

1.2

1.5

7.4

6.8

1781-85

6.9

3.9

0.8

0.7

2.7

6.4

8.2

1785

1.0

6.5

2.0

0.0

1.5

18.0

1 W Westerly, NW North Westerly, N Northerly, E Easterly, S Southerly, A Anticyclonic, C Cyclonic.

Thus there is a striking similarity (blocked and very cyclonic) in the PSCM "signatures" of 1816
and 1785.

As rainfall over England and Wales has been found to be closely correlated with the C-index, it
is not surprising that very heavy falls of rain occurred over the region in July 1816 (Figure 6).
The map, however, shows that it was not uniformly wet over the British Isles or continental
Europe. For instance, while rainfall over southwestern Ireland, southern Wales, southwestern
England, most of France, parts of Belgium, Holland and western Germany exceeded 200% of
normal, northwestern Scotland, Orkney, Shetland, Denmark, Norway and Italy were drier than
usual. Contemporary accounts confirm this contrasting pattern of wet and dry regions:

Europe

Melancholy accounts have been received from all parts of the Continent of the
unusual wetness of the season; property in consequence swept away by
inundation, and irretrievable injuries done to the vine yards and corn crops.
In several provinces of Holland, the rich grass lands are all under water, and
scarcity and high prices are naturally apprehended and dreaded. In France, the
interior of the country has suffered greatly from the floods and heavy rains.

"The Norfolk Chronicle", 20 July 1816

Ireland

With depressions centred over or near Ireland for most of the month, the weather over the
country was very unsettled and wet. Apparently, all parts had more rain than usual, with the
extreme southwest probably having more than twice the normal amount (Figure 6).

The summer and autumn were excessively wet and cloudy. ... the sun was in
general obscured by clouds during the months of July, August and September.

Great thunderstorms occurred during the month of July, accompanied with
hail of an unusually large size. These storms were general throughout the
country.

368

July - wet, great storms, and inundations in England and Scotland, as well as
throughout this country ... The month was, without, perhaps, the exception
of a single day, a continuity of showers of hail or rain, and at the same time
very cold.

Snow remained on some of the hills in Scotland until the middle of July,
during which month great thunderstorms occurred in England.

In consequence of the incessant rain, there is a great blight in the wheat crop,
particulary in Wicklow and Tipperary: the rain was so severe that scarcely
any corn was left standing. For many years so untoward a season had not
been experienced, not one week of fine weather since May. Eight weeks of
rain in succession. Hay and corn crops in a deplorable state. The grains of
corn in many places are covered with a reddish powder like rust, which has
proved very destructive to the crop, especially in the counties of Kilkenny and
Antrim.1 The wheat crop was especially injured. Great floods occurred in the
Boyne.

The fields of corn presented a lamentable appearance, in many places being
quite black. Before the crop was reaped, re-vegetation had commenced, and
green shoots were perceived on the fields.

The harvest of grain was uncommonly late both in this country and in
England; corn remained uncut during the latter parts of October and
November, and much of it was altogether lost. The cold of this season proved
highly injurious to the crop of potatoes also. These, which constitute the
principal or only food of the poor in most parts of the country, were small
and wet, and probably more defective in nutriment than the grain.

The potato crop both in England and Scotland was defective.

"The Census of Ireland," 1851

Denmark

This month for the most part good weather. Quite warm 21° and frequent
rain, although this did not do any harm. On the other hand in Germany and
Switzerland terrible damage occurred with rivers flooding. This was caused
by persistent rain ... whole tracts of land were under water. The hay harvest
was also ruined in England.

"A Jutland Weather Diary" (Ribe)

This may have been the result of volcanic aerosol particles being washed out of the atmosphere by the rain which,
in turn, might have been intensified by the increase in condensation nuclei. Editor's note: Perhaps the possibility
of fungal rust should be considered also.

369

Western Russia and the Baltic Sea Coast

The city of St. Petersburg [Leningrad] has for a month past suffered by
drought and prayers for rain have been offered up at Riga and Dantzig while
Germany is devastated by inundations and the churches of Paris are filled with
suppliants praying the Almighty for dry weather.

"Records of the Seasons"

Conclusion

One of the main objectives of this book has been to determine to what extent the Tambora
eruption in 1815 affected world climate. Already we know that some mid-latitude regions of the
northern hemisphere, such as eastern North America and western Europe, were much cooler than
normal in the following year, 1816. There is an interesting parallel in the 1780s when it is
estimated that annual mean temperatures in mid-latitudes fell by 1.3°C after eruptions in Iceland
and Japan in 1783. However, there appear to be two major points of difference: the timing and
length of cooling. By all accounts it appears that, unlike 1815-16, the cooling signal in the mid-
1780s was strongest not in the year immediately following the eruption but in 1785, two years
after the event. Nevertheless I have shown that there were some notable similarities in the
circulation patterns of the two cold years, 1816 and 1785. Furthermore, the marked increase in
cyclonicity over the British Isles in July 1816 is in accordance with Lamb's (1977) finding that
there is a tendency for the subpolar low-pressure zone (the "Icelandic Low") to be displaced
southwards over the British Isles during the first July after a great eruption, resulting typically
in cold wet summers over the region. Another area of cyclonic activity near Newfoundland gave
similar weather conditions over eastern North America. However, the volcanic signal apparently
soon died away, with temperatures recovering to above normal values by 1818. On the other
hand, after high-latitude eruptions (e.g., those of 1783), pressure and related temperature
anomalies in mid-latitudes appear to persist longer - the circulation patterns determined for July
in the cold year of 1785 confirm this trend.

Acknowledgements

Drs. A.E.J. Ogilvie and P.D. Jones kindly helped in processing various historical weather data
from the archives of the Climatic Research Unit. Observations from Dublin and France were
kindly supplied by Dr. J.G. Tyrrell (University College, Cork) and Dr. D. Hubert (Observatoire
de Meudon), respectively.

References

Baker, T.H. 1883. Records of the Seasons, Prices of Agricultural Produce and Phenomena
Observed in the British Isles. Simpkin, Marshall and Co., London.

Dublin. 1856. The Census of Ireland for the Year 1851. H.M.S.O., London.

Jones, P.D., T.M.L. Wigley and K.R. Briffa. 1987. Monthly mean pressure reconstructions for
Europe (back to 1780) and North America (to 1858). DOE Technical Report No. 37,
United States Department of Energy, Carbon Dioxide Research Division, Washington,
D.C.

370

Kington, J. 1988. The Weather of the 1780s Over Europe. Cambridge University Press,
Cambridge.

Lamb, H.H. 1972. British Isles weather types and a register of the daily sequence of circulation
patterns, 1861-1971. Geophysical Memoirs No. 116. H.M.S.O., London.

. 1977. Climate: Present, Past and Future, Volume 2, Climatic History and the Future.

Methuen, London.

Ludlam, F.H. 1966. The Cyclone Problem: A History of Models of the Cyclonic Storm. Imperial
College of Science and Technology, London.

Murray, R. and R.P.W. Lewis. 1966. Some aspects of the synoptic climatology of the British
Isles as measured by simple indices. Meteorological Magazine 95:193-203.

Von Rudloff, H. 1967. Die Schwankungen und Pendelungen des Klimas in Europa seit dem
Beginn der regelmilssigen Istrumenten-Beobachtungen (1670). Vieweg, Braunschweig.

371

The Climate of Europe during the 1810s with Special
Reference to 1816

K.R. Briffa1 and P.D. Jones1
Abstract

The long climatic records available for Europe are used to place the seasonal temperature,
precipitation and sea-level pressure anomaly maps for 1816 into their longer-term context. The
prevailing climate of the decade of the 1810s (1810-19) is also described with reference to
modern climatic normals. The 1810s were probably one of the coldest decades recorded over
Europe since comparable records began about 1750. It was only the weather during the spring
and, more particularly, the summer of 1816 that was highly anomalous with respect to both
recent normals and those for the 1810s.

Tree-ring-based reconstructions of temperature for a 'summer' (April-September) season are
available in the form of anomaly maps back to 1750. They indicate that the summer of 1816 was
the coldest since 1750 in Britain, that it was the second coldest (after 1814) in central Europe and
that in Scandinavia conditions were near normal.

Introduction

Many studies have considered the weather extremes that occurred during the summer of 1816,
the so-called "year without a summer" (Landsberg and Albert 1974; Stommel and Stommel
1979). Studies have tended to concentrate on the particular season itself, rather than considering
the weather and climate of the rest of 1816 and the decade of the 1810s.

In this article we propose to make use of the long records of temperature, precipitation and mean
sea-level pressure (MSLP) available for most of Europe. We will describe seasonal anomaly maps
for 1816 with respect to twentieth century reference periods and in relation to those of the 1810s
(defined here as 1810-19). We also compare the climate of the 1810s to recent reference periods.

Finally, previously published maps of mean April-September temperature reconstructed from a
network of maximum-latewood-density tree-ring chronologies in Europe are reproduced for each
of the years 1810-19.

Data

Instrumental recording of air temperature and precipitation totals extends back in Europe to the
late seventeenth century. Most of the pre-twentieth century data have been assembled in computer
compatible form in data archives. Here we use the compilation of air temperature and
precipitation data produced by Bradley et al. (1985). This archive contains temperature data for
46 stations in Europe with series that extend over most of the years of the 1810s (Table 1;
Figure 1). Of these 46 stations, 12 do not have comparable data through to and encompassing the
twentieth century. We can still use these more restricted data, however, to compare the average
temperature of 1816 to that of the 1810s.

Climatic Research Unit, University of East Anglia, Norwich NR4 7TJ, U.K.

372

Table 1: Names and Locations of Stations with Temperature Data for the 1810s
Continuous to the Present Day.

Lat.(°N) Long.

Trondheim

64.3

10.5E

Stockholm

59.4

18.1E

Tomeo

66.4

23. 8E

Woro

63.2

22. OE

Gordon Castle

57.6

3.1W

Edinburgh

55.9

3.2W

Manchester

53.4

2.3W

Greenwich

51.5

Copenhagen

55.7

12.6E

10.

De Bilt

52.1

5.2E

11.

Basel

47.6

7.6E

12.

Geneva

46.2

6.2E

13.

Montdidier

49.7

2.6E

14.

Chalons

48.9

4.4E

15.

Paris

48.8

2.5E

16.

Strasbourg1

48.6

7.6E

17.

Nice

43.7

7.2E

18.

Berlin

52.5

13.4E

19.

Karlsruhe1

49.0

8.4E

20.

Stuttgart1

48.8

9.2E

21.

Regensberg1

49.0

12.1E

22.

Augsburg1

48.4

10.4E

23.

Munchen1

48.1

11.7E

24.

Hohenpei ssenberg 1

47.8

11. OE

25.

Kremuenster1

48.1

14. IE

26.

Wien Hohe Warte

48.2

16.4E

27.

Innsbruck

47.3

11. 4E

28.

Klagenfurt

46.7

14.3E

29.

Prague

50.1

14.3E

30.

Leobschutz

50.2

17.8E

31.

Gdansk

54.4

18.6E

32.

Warsaw

52.2

21. OE

33.

Wroclaw

51.1

17.0E

34.

Budapest

47.5

19.0E

35.

Udine

46.0

13.1E

36.

Turin

45.2

7.7E

37.

Milan

45.4

9.2E

38.

Padua

45.4

12. OE

39.

Bologna

44.5

11. 5E

40.

Rome

41.7

12.5E

41.

Palermo

38.1

13.4E

42.

Arkhangel

64.6

40.6E

43.

Leningrad

60.0

30.3E

44.

Vilnjus

54.6

25.3E

45.

Kazan

55.8

49. IE

46.

Kiev

50.5

30.5E

Not labelled on Figure I.

373

374

Although the Bradley et al. (1985) compilation contains details of the sources of the data and the
methods, where known, by which the observations were made, it does not consider the long-term
homogeneity of the individual station data sets. The homogeneity of the station temperature data
used here has, however, been assessed by Jones et al. (1985).

For precipitation, the Bradley et al. (1985) compilation contains data for 29 sites covering the
1810s. Assessment of the homogeneity of the precipitation data is a considerably more difficult
task than for air temperature. Although the stations used here have not been assessed for
homogeneity, the data for 27 of the sites are among the 180 or so homogeneous European
precipitation records assembled by Tabony (1980, 1981). Only the data for Warsaw and Prague
are not in this set. The locations of the 29 precipitation sites we have used are shown (Figure 2;
Table 2).

Some of these early temperature and precipitation series have been used in conjunction with early
station pressure records by Jones et al. (1987) to reconstruct gridded monthly-mean mean sea-
level pressure values (MSLP) over Europe extending back to 1780. Jones et al. used a principal
components regression technique that involves fitting equations expressing MSLP at individual
grid points in terms of pressure, temperature and precipitation series at all stations in the
predictor network. The fitting was carried out over a 75-year (1900-74) 'calibration' period, and
the reliability of the gridded reconstructions was assessed by comparing the estimated data with
actual observations over an independent 'verification' period, 1873-99. Jones et al. (1987)
showed that over Europe, between 65-40 °N and 10oW-30°E, the reconstructions are of high
quality with 80% or more of the variance of the observed pressure data being explained in each
of the separate monthly reconstructions.

From this bank of reconstructed pressures we have extracted the data for individual months and
averaged them to produce maps of MSLP anomalies for the four standard seasons of the year
1816.

Anomaly Maps

Figure 3 shows seasonal temperature anomaly maps for 1816 with respect to the reference period
1951-70. Winter in this and subsequent figures is taken to be December 1815 to February 1816.
All four seasons are shown to have been generally cooler in 1816 compared with the reference
period. Warmer conditions were experienced only over northern Mediterranean coasts and
European parts of the Soviet Union, and then only in spring, summer and autumn. The most
anomalously cool regions were Scandinavia and northern British Isles (during winter, spring and
autumn) and central Europe (in summer).

Figure 4 shows seasonal precipitation anomaly maps for 1816 (expressed as percentages of the
1921-60 reference period). Most regions of western Europe were drier than normal except for
summer. Below normal precipitation is evident over central and southern Europe in winter and
to some extent in spring and autumn. During summer the only relatively dry areas were southern
Italy and northwestern Scotland.

375

Figure 2: Location of the 29 sites with precipitation-gauge data during the 1810s. Details of the sites are
given in Table 2 (from Bradley et al. 1985).

376

Table 2: Names and Locations of Stations with Precipitation Data for the 1810s Continuous
to the Present Day.

Lat.(°N)

Long.

1 T T 1 _

1 . Uppsala

59.9

IT £LT*

17.6b

2. Lund

33. /

13.2b

3. Inverness

3 I .J

4.ZW

4. ha 11 anus

JJ.O

o.zW

5. hdinburgh

55.5

3.2W

6. Manchester

Z. J w

7. Mansfield

53. 1

1.1W

8. Podehole

52.8

0. 1W

9. Kew

51.5

0.3W

1U. Uxtord

CI T

51. /

l.ZW

11. Hoofdoorp

O 1
JZ.J

4. /il

11 T -11 ^

12. Lille

JV.O

3. lh

13. Montaiaier

Ann
49. /

2.ob

14. rans

A 0 Q

Z.Db

15. Nancy

4o. /

o.zb

16. Strasbourg

48.0

/.ob

i/. La i\ocneue

Afi 1
10. 1

1 1 w
1 . 1 w

18. Toulouse

A1 A

1 /1C

19. Marseille

4J.3

5.4b

zU. 1 ner

49.0

6.7b

21. Karlsruhe

49.0

8.4b

22. Klagenfurt

46.7

14.3b

23 . Prague

jU. 1

14. Jb

24. Warsaw

52.2

21.0b

25. Udine

46.0

13.1b

26. Milan

45.4

9.2E

27. Padua

45.4

12.0E

28. Bologna

44.5

11. 5E

29. Rome

41.7

12.5E

In Figures 5-8 we show similar seasonal anomaly maps for temperature and precipitation, placing
the 1810s in the context of modern reference periods, and 1816 in relation to the 1810s. The
1810s (Figure 5) were colder than the recent reference period during winter and autumn but were
somewhat milder during spring and summer, particularly over eastern and southern Europe. The
relative coolness of the 1810s with respect to 1951- 70 in winter and spring means that 1816 was
less anomalous when viewed against this decade as a whole (Figure 7). For precipitation, the
1810s were generally drier than the 1921-60 reference period (Figure 6), other than over the
British Isles and Scandinavia in spring and Italy during summer and autumn.

377

Figure 3: Seasonal temperature anomaly maps in degrees Celsius for 1816 with respect to the 1951-70
reference period. Winter is the average for December 1815 to February 1816. Spring
(March-May), summer (June-August) and autumn (September-November).

378

Figure 4: Seasonal precipitation departures for 1816: values expressed as percentages of the 1921-60
reference period mean.

379

Figure 5: Seasonal temperature anomaly maps for the 1810s (1810-19) with respect to the 1951-70
reference period.

380

Figure 6: Seasonal precipitation departures for the 1810s: values expressed as percentages of the 1921-60
reference period mean.

381

Figure 7: Seasonal temperature anomaly maps for 1816 with respect to the average for the 1810s
(1810-19).

382

383

Comparison of the 1816 Seasonal Temperature, Precipitation and MSLP Anomaly Maps
Winter (Figure 9)

During the winter of 1815-16 all of Europe was affected by anomalously low pressure with
respect to the 1941-70 reference period. Both southern and western Europe were affected by
greater advection from eastern Europe, which would tend to bring drier and cooler conditions to
these regions. Enhanced northerly circulation over western Europe resulted in below normal
precipitation in all areas except north-facing coasts.

Spring (Figure 10)

Europe is again shown to be almost entirely under the influence of lower-than-normal pressure.
The negative pressure anomaly in this season is centred over northern France, implying that both
Britain and Scandinavia experienced increased easterly and northeasterly weather, from Finland
and the Gulf of Bothnia, leading to cold temperatures. Drier conditions over continental Europe
may have resulted from the stagnation of a number of depressions over this region. Wetter than
normal weather over northern Europe was associated with a greater degree of air flow over
adjacent seas.

Summer (Figure 11)

Virtually the whole of Europe was affected by anomalously low pressure centred over northern
Germany and Denmark. Milder conditions prevailed over European parts of the Soviet Union
because of the influence of increased southerly flow across these areas. Britain and the rest of
western Europe were affected by anomalous northerly and northwesterly airflow bringing cooler
temperatures. The coldest conditions of the summer occurred in northern Alpine regions. Over
Scandinavia, in contrast to winter and spring, conditions were near to the recent normal.
Precipitation was considerably greater over northern France and southern England.

The summer (June-August) of 1816 was the coldest recorded in the Central England temperature
series (Manley 1974; updated in Jones 1987). Temperatures were 2.2°C colder than the 1931-60
average. The Manley series extends back to 1659 (though with slightly lower reliability before
1721).

Autumn (Figure 12)

Again most of Europe was under the influence of anomalously low pressure, although less intense
than in the other seasons. The centre of the anomaly was located over Poland, whereas pressure
was near normal over Ireland and Scotland. Enhanced northerly and northeasterly air circulation
over Scandinavia, Britain and central Europe, particularly north of the Alps would have led

Internet Archive Audio

Featured

Top

Images

Featured

Top

Software

Featured

Top

Books

Featured

Top

Video

Featured

Top

Mobile Apps

Browser Extensions

Archive-It Subscription

Save Page Now

Full text of "The year without a summer? : world climate in 1816"

See other formats